MLXID/DeepLearning
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

reduce ram usage

Browse Source 
Co-authored-by: Copilot <copilot@github.com>
This commit is contained in:
xie_x1
2026-06-24 10:12:15 +02:00
co-authored by Copilot
parent 1891f141ea
commit b728668440
1 changed files with 33 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/etaInterpolationFromPoints.py
+33 -14
View File
@@ -6,22 +6,41 @@ import numpy as np
NEtaBins = 201
def interpolate_eta_from_points(points):
points_integer = np.int32(points)
points_fractional = points - points_integer
points_fractional_x = points_fractional[:, 0]
points_fractional_y = points_fractional[:, 1]
def interpolate_eta_from_points(points, chunk_size=5_000_000):
print('[EtaInterpolation]: Building eta histogram from points (Chunked)...')
eta_hist2D = np.zeros((NEtaBins, NEtaBins), dtype=np.float64)
print('[EtaInterpolation]: Building eta histogram from points...')
eta_hist2D = np.histogram2d(
points_fractional_x, points_fractional_y,
bins = NEtaBins, range=[[0, 1], [0, 1]]
)[0] ### shape (201, 201), X, Y, according build_xy
# --- building eta histogram in chunks ---
for i in range(0, len(points), chunk_size):
start_idx = i
end_idx = min(i + chunk_size, len(points))
chunk = points[start_idx:end_idx]
frac_x = chunk[:, 0] - np.int32(chunk[:, 0])
frac_y = chunk[:, 1] - np.int32(chunk[:, 1])
hist_chunk, _, _ = np.histogram2d(
frac_x, frac_y,
bins=NEtaBins, range=[[0, 1], [0, 1]]
)
eta_hist2D += hist_chunk
U_tab, V_tab = build_xy_lut_Rosenblatt(eta_hist2D)
print('[EtaInterpolation]: Performing bilinear interpolation...')
x_frac, y_frac = bilinear_xy_lookup(points_fractional_x, points_fractional_y, U_tab, V_tab)
print('[EtaInterpolation]: Performing bilinear interpolation (Chunked)...')
corrected_points = np.empty_like(points)
# --- performing bilinear interpolation in chunks ---
for i in range(0, len(points), chunk_size):
chunk = points[i:i + chunk_size]
int_x = np.int32(chunk[:, 0])
int_y = np.int32(chunk[:, 1])
frac_x = chunk[:, 0] - int_x
frac_y = chunk[:, 1] - int_y
x_frac_corr, y_frac_corr = bilinear_xy_lookup(frac_x, frac_y, U_tab, V_tab)
corrected_points[i:i + chunk_size, 0] = int_x + x_frac_corr
corrected_points[i:i + chunk_size, 1] = int_y + y_frac_corr
corrected_points = points_integer + np.stack([x_frac, y_frac], axis=-1)
return corrected_points