Model I Setup - Broken Links, Hardcoded Paths, and Import Errors

[BeathovenGala]

Hi, I haven't been able to access the datasets for the Deeplense project as I was trying to test a few things

The problem is that the Google Drive links in DeepLenseSim/Model_I/README.md seem to be down (both return 404):

• Training data — 404
• Testing data — 404

Since I can’t download the dataset right now, I wanted to check:

• Are there updated links available?
• Or is there another recommended way to get the Model I data?

Let me know what the best approach is. Till then, I'll work with the scripts and generate the dataset locally.

[BeathovenGala]

Technical Issues While Simulating Model I Data (and Fixes Applied)

As the dataset links were unavailable, I proceeded to generate the Model I data locally using the simulation scripts.
Although I ran into a few other issues,I saw PRs related to the similar issues in the other models too

1. ImportError with pyHalo

The deeplense/lens.py script fails to import CDM due to an outdated import path.

Error:

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

Fix:

# Old
from pyHalo.preset_models import CDM

# New (works with latest pyHalo)
from pyHalo.PresetModels.cdm import CDM

---

2. Hardcoded Absolute Paths

sim_no_sub.py, sim_axion.py, and sim_cdm.py include hardcoded save paths such as:

/users/mtoomey/scratch/...

Fix — switched to relative paths:

import os

base_dir = os.path.dirname(os.path.abspath(__file__))
save_dir = os.path.join(base_dir, '../data/Model_I/no_sub/')
os.makedirs(save_dir, exist_ok=True)

np.save(save_dir + 'no_sub_sim_' + str(random.getrandbits(128)), File)

---

3. Numpy Array Shape Error in sim_axion.py

Combining [image array, scalar] directly caused an inhomogeneous shape error.

Error:

ValueError: setting an array element with a sequence...

Fix:

File = np.array([lens.image_real, axion_masses[i]], dtype=object)

---

4. Missing Dependencies

Had to install/update the following:

• astropy
• mcfit
• (Ensured lenstronomy was up to date)

pip install astropy mcfit lenstronomy --upgrade

---

Should I proceed with similar cleanup and simulation for Models II, III, and IV, or will the dataset links be restored soon?