Skip to content

Fix: Implement path portability and add default resolution to train.py and data.py #112

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Shrishagk wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Fix: Implement path portability and add default resolution to train.py and data.py #112

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f995f4b
Fix path portability and add default resolution in train.py
Shrishagk Jan 26, 2026
16573fd
Improve path handling in data.py using os.path.join
Shrishagk Jan 26, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 19 additions & 8 deletions Grid_based_strong_lensing_for_unsupervised_super_resolution_Anirudh_Shankar/data.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
import torch
import numpy as np

import os
class LensingDataset(torch.utils.data.Dataset):
def __init__(self, directory, classes, num_samples):
"""
Expand All @@ -20,15 +20,26 @@ def __len__(self):
"""
return self.num_samples*len(self.classes)





def __getitem__(self, index):
"""
Supplies LR images

:param index: Index in the dataset to look for
:return: LR image, min-max normalized
"""
selected_class = self.classes[index//self.num_samples]
class_index = index%self.num_samples
image = torch.tensor(np.array([np.load(self.directory+selected_class+'/sim_%d.npy'%(class_index))]))
image = (image - torch.min(image))/(torch.max(image)-torch.min(image))
selected_class = self.classes[index // self.num_samples]
class_index = index % self.num_samples

# USE os.path.join for portability
file_path = os.path.join(self.directory, selected_class, 'sim_%d.npy' % class_index)

image = torch.tensor(np.array([np.load(file_path)]))

# Small safety check: handle division by zero if image is blank
img_min = torch.min(image)
img_max = torch.max(image)
if img_max > img_min:
image = (image - img_min) / (img_max - img_min)

return image
32 changes: 18 additions & 14 deletions Grid_based_strong_lensing_for_unsupervised_super_resolution_Anirudh_Shankar/train.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -55,8 +55,8 @@ def parse_args():
help='weight of the vdl loss')

# Performance / architecture options
parser.add_argument('--resolution', type=float,
help='arcsecond per pixel resolution the images are captured in')
parser.add_argument('--resolution', type=float, default=0.1,
help='arcsecond per pixel resolution the images are captured in')
parser.add_argument('--magnification', type=int, default=2,
help='magnification value achieved by the SR network')
parser.add_argument('--n-mag', type=int, default=1,
Expand Down Expand Up @@ -137,18 +137,22 @@ def parse_args():
'hyperparameters',
'|param|value|\n|-|-|\n%s'%('\n'.join([f'|{key}|{value}' for key, value in vars(args).items()])),
)

# --- load precomputed sparse mappings and maps --------------------------
# These files must exist in working dir. They are moved to args.device below.
cross_grid_to_log = torch.load('scatter_to_log_128.pt').to(args.device)
cross_grid_forward_from_log = torch.load('forward_from_log_128.pt').to(args.device)
cross_grid_from_log = torch.load('scatter_from_log_128.pt').to(args.device)
cross_grid_backward = torch.load('sparse_grid_fracs_euclid_backward.pt').to(args.device)

# convergence maps
source_convergence_map = torch.load('source_convergence_map.pt').to(args.device)
image_convergence_map = torch.load('image_convergence_map.pt').to(args.device)

# Get the script directory
script_dir = os.path.dirname(os.path.abspath(__file__))

# Define the folder where the grids are kept
grid_dir = os.path.join(script_dir, 'grid_matrices')

# Load the files from that specific folder
cross_grid_to_log = torch.load(os.path.join(grid_dir, 'scatter_to_log_128.pt')).to(args.device)
cross_grid_forward_from_log = torch.load(os.path.join(grid_dir, 'forward_from_log_128.pt')).to(args.device)
cross_grid_from_log = torch.load(os.path.join(grid_dir, 'scatter_from_log_128.pt')).to(args.device)
cross_grid_backward = torch.load(os.path.join(grid_dir, 'sparse_grid_fracs_euclid_backward.pt')).to(args.device)

# Check if these maps are also in grid_matrices or the root
source_convergence_map = torch.load(os.path.join(grid_dir, 'source_convergence_map.pt')).to(args.device)
image_convergence_map = torch.load(os.path.join(grid_dir, 'image_convergence_map.pt')).to(args.device)

# --- PSF kernel setup ---------------------------------------------------
# gaussian_kernel returns (Z, X, Y) as numpy arrays in the previous file.
# Converting to torch.tensor is fine but be mindful of dtype/device.
Expand Down