feat: Add GPU-accelerated operations via PyTorch#21

Closed

maragall wants to merge 18 commits into

Cephla-Lab:mainfrom

maragall:feature/gpu-acceleration

Commits on Mar 7, 2026

feat: Add GPU-accelerated operations via PyTorch

authored and
maragall
committed
style: Apply black formatting

hongquanli
authored and
maragall
committed
fix: Remove unused variables in _match_histograms_torch

hongquanli
authored and
maragall
committed
feat: Add dtype preservation to shift_array, match_histograms, block_reduce

hongquanli
authored and
maragall
committed
fix: Handle 2D block_size for 3D arrays in _block_reduce_torch

hongquanli
authored and
maragall
committed
refactor: Extract duplicate data_range calculation in compute_ssim

hongquanli
authored and
maragall
committed
refactor: Add named constants for magic numbers (_FFT_EPS, _SSIM_K1, _SSIM_K2)

hongquanli
authored and
maragall
committed
test: Add CPU fallback and dtype preservation tests

hongquanli
authored and
maragall
committed
test: Add subpixel phase correlation tests

hongquanli
authored and
maragall
committed
refactor: Add _PARABOLIC_EPS constant for subpixel refinement

hongquanli
authored and
maragall
committed
fix: Guard against 1-pixel arrays in _shift_array_torch

hongquanli
authored and
maragall
committed
refactor: Add __all__ export list and document legacy compatibility vars

hongquanli
authored and
maragall
committed
test: Refactor tests to use rng fixture and pytest class style

hongquanli
authored and
maragall
committed
chore: Clean up unused imports and variables

hongquanli
authored and
maragall
committed
Add type hints to public functions in utils.py

authored and
maragall
committed
Fix histogram matching bug and improve type hints

authored and
maragall
committed
Cosmetic cleanups and fix shift_array CPU/GPU consistency

authored and
maragall
committed
Document GPU path placeholder values in phase_cross_correlation

authored and
maragall
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: Add GPU-accelerated operations via PyTorch#21

feat: Add GPU-accelerated operations via PyTorch#21
maragall wants to merge 18 commits into
Cephla-Lab:mainfrom
maragall:feature/gpu-acceleration

Commits on Mar 7, 2026

feat: Add GPU-accelerated operations via PyTorch

style: Apply black formatting

fix: Remove unused variables in _match_histograms_torch

feat: Add dtype preservation to shift_array, match_histograms, block_reduce

fix: Handle 2D block_size for 3D arrays in _block_reduce_torch

refactor: Extract duplicate data_range calculation in compute_ssim

refactor: Add named constants for magic numbers (_FFT_EPS, _SSIM_K1, _SSIM_K2)

test: Add CPU fallback and dtype preservation tests

test: Add subpixel phase correlation tests

refactor: Add _PARABOLIC_EPS constant for subpixel refinement

fix: Guard against 1-pixel arrays in _shift_array_torch

refactor: Add all export list and document legacy compatibility vars

test: Refactor tests to use rng fixture and pytest class style

chore: Clean up unused imports and variables

Add type hints to public functions in utils.py

Fix histogram matching bug and improve type hints

Cosmetic cleanups and fix shift_array CPU/GPU consistency

Document GPU path placeholder values in phase_cross_correlation