**Title:** [Bug] Hardcoded absolute paths in Model_I/II simulation scripts

[Shrishagk]

Issue: Hardcoded Absolute File Paths Break Cross-Platform Execution

Description

Currently, several simulation scripts (for example, Model_I/sim_axion.py) contain hardcoded absolute file paths pointing to a specific user’s directory on a Linux cluster (e.g., /users/mtoomey/scratch/...). This causes the scripts to crash immediately on Windows, macOS, or any system other than the original author's.

This causes the scripts to crash immediately on Windows, macOS, or any system other than the original author’s environment.

---

Steps to Reproduce

1. Clone the repository on a Windows or macOS machine.
2. Attempt to run the script:

   python Model_I/sim_axion.py

3. The script fails with a FileNotFoundError because the directory /users/mtoomey/... does not exist.### Expected Behavior
   The scripts should use relative paths or a configurable base directory so that data is saved within the project folder regardless of the operating system.

---

Proposed Fix

Replace absolute path strings with os.path joins to the current working directory.
Python

import os
base_path = os.path.join(os.getcwd(), 'data', 'Model_I')
os.makedirs(base_path, exist_ok=True)

[Shrishagk]

Hello maintainers,

I’ve investigated the hardcoded path issue in the Model_I simulation scripts. I've identified that the absolute paths to the cluster directories cause crashes on local Windows/macOS environments.

I have a fix ready using os.path.join and os.getcwd() to ensure the data is saved in a portable, relative directory structure.

[srividya-0001]

Hi @Shrishagk, thank you for investigating the issue and sharing the update.
Have you opened a PR for the fix? If yes, please share the PR link — I can test it on Windows to verify cross-platform behavior.
If not, I can proceed with implementing the fix and submit a PR.

[Shrishagk]

mwt5345/DeepLenseSim#1

…

On Mon, 19 Jan 2026 at 21:11, Srividya Kollepu ***@***.***> wrote: *srividya-0001* left a comment (ML4SCI/DeepLense#99) <#99 (comment)> Hi @Shrishagk <https://github.com/Shrishagk>, thank you for investigating the issue and sharing the update. Have you opened a PR for the fix? If yes, please share the PR link — I can test it on Windows to verify cross-platform behavior. If not, I can proceed with implementing the fix and submit a PR. — Reply to this email directly, view it on GitHub <#99 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BTFHXSCQA2CTCHEWVQFGWIL4HT3LVAVCNFSM6AAAAACSFIXGZOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTONRYHE2DSMBYGM> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[srividya-0001]

Thanks for sharing the PR link, @Shrishagk!
I’ll review the changes and test it on Windows. I’ll update here with results / any issues I find.

[srividya-0001]

Hi @Shrishagk, thanks for sharing the PR and working on this! 🙌
I tried testing the scripts on Windows (fresh venv, Python 3.11).

At the moment, Model_I/sim_axion.py fails early due to a dependency import issue:

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

So I couldn’t reach full runtime execution yet. This seems separate from the hardcoded path issue.

Also, while checking the code, I noticed that hardcoded output paths are still present in a few scripts. For example in Model_I/sim_axion.py:

np.save('/users/mtoomey/scratch/deeplense/Model_I_test/axion/axion_sim_' + str(random.getrandbits(128)), File)

Similar /users/.../scratch/... occurrences appear in the other simulation scripts too.
Maybe we can update those to portable relative/configurable paths (Path.cwd / output_dir), and then the issue should be fully resolved.

Happy to help test again on Windows once updated! 🙂

[srividya-0001]

Hi @Shrishagk, I opened a separate issue for the pyHalo import mismatch that blocks end-to-end testing on Windows: #100.
This looks separate from the hardcoded path fix, but it currently prevents the scripts from running fully during testing

[Shrishagk]

Hi @srividya-0001 , Thanks for going deep into it. I will try to update all the hardcoded paths which are still present so that the issue can be clearly solved.

[srividya-0001]

Update (Windows Testing): I re-tested the latest changes on Microsoft Windows 10 Pro using a fresh Python 3.11 virtual environment.

• The hardcoded absolute path issues reported in this thread are resolved via PR: Fix hardcoded absolute paths in simulation scripts mwt5345/DeepLenseSim#1

• The pyHalo CDM import/API mismatch that was blocking end-to-end execution is also resolved (version-agnostic import logic).

• End-to-end execution now proceeds successfully — Model_I/sim_axion.py completes the full run (5000 samples generated).

Overall, the changes significantly improve portability and allow full script execution on Windows/local setups. Happy to help with any further testing or debugging if needed.