Merge pull request #14 from esa/Rescaled-sampling-in-frequency-domain…

…-(to-have-less-volatile-eps-vs-freq)

Rescaled sampling in frequency domain (to have less volatile eps vs freq)

nidn/training/model/model_to_eps_grid.py

@@ -19,14 +19,17 @@ def _eval_model(model, Nx_undersampled, Ny_undersampled, N_layers, target_freque
     y = torch.linspace(-1, 1, Ny_undersampled)
     z = torch.linspace(-1, 1, N_layers)
 
+    # Scales sampling domain of frequencies
+    freq_scaling = 32.0
+
     # Linearly spaced frequency points
-    #     freq = torch.linspace(-1,1,len(run_cfg.target_frequencies))
+    # freq = torch.linspace(-freq_scaling, freq_scaling, len(target_frequencies))
 
     # Logspaced frequency points
     # Normalize to max val = 1
     freq = torch.tensor(target_frequencies / max(target_frequencies))
-    # Transform to -1 to 1
-    freq = (freq * 2) - 1
+    # Transform to -scaling to scaling
+    freq = (freq * 2 * freq_scaling) - freq_scaling
 
     # Create a meshgrid from the grid ticks
     X, Y, Z, FREQ = torch.meshgrid((x, y, z, freq))
@@ -76,8 +79,7 @@ def _regression_model_to_eps_grid(model, run_cfg: DotMap):
 
     # Initialize the epsilon grid
     eps = torch.zeros(
-        [run_cfg.Nx, run_cfg.Ny, run_cfg.N_layers, run_cfg.N_freq],
-        dtype=torch.cfloat,
+        [run_cfg.Nx, run_cfg.Ny, run_cfg.N_layers, run_cfg.N_freq], dtype=torch.cfloat,
     )
 
     # Net out is [0,1] thus we transform to desired real and imaginary ranges

nidn/trcwa/init_trcwa.py

@@ -25,13 +25,7 @@ def _init_trcwa(eps_grid, target_frequency):
 
     # Initialize TRCWA object
     trcwa = TRCWA(
-        TRCWA_NG,
-        TRCWA_L1,
-        TRCWA_L2,
-        freqcmp,
-        TRCWA_THETA,
-        TRCWA_PHI,
-        verbose=0,
+        TRCWA_NG, TRCWA_L1, TRCWA_L2, freqcmp, TRCWA_THETA, TRCWA_PHI, verbose=0,
     )
 
     # Add vacuum layer at the top

setup.py

@@ -0,0 +1,54 @@
+"""A setuptools based setup module.
+See:
+https://packaging.python.org/guides/distributing-packages-using-setuptools/
+https://github.com/pypa/sampleproject
+"""
+
+# Always prefer setuptools over distutils
+from setuptools import setup, find_packages
+
+setup(
+    name="nidn",
+    version="0.1.0",
+    description="A package for inverse material design of nanostructures using neural networks.",
+    long_description=open("README.md").read(),
+    long_description_content_type="text/markdown",
+    url="https://github.com/esa/nidn",
+    author="ESA Advanced Concepts Team",
+    author_email="pablo.gomez@esa.int",
+    install_requires=[
+        "dotmap>=1.3.24",
+        "loguru>=0.5.3",
+        "matplotlib>=3.3.3",
+        "numpy>=1.20.0",
+        "pandas>=1.3.1",
+        "scipy>=1.6.0",
+        "tqdm>=4.56.1",
+        "torch>=1.9",
+    ],
+    classifiers=[
+        "Development Status :: 3 - Alpha",
+        "Intended Audience :: Developers",
+        "Intended Audience :: Science/Research",
+        "Topic :: Scientific/Engineering :: Mathematics",
+        "License :: OSI Approved :: GNU General Public License v3 (GPLv3)",
+        "Programming Language :: Python :: 3.8",
+    ],
+    packages=[
+        "nidn",
+        "nidn.materials",
+        "nidn.plots",
+        "nidn.tests",
+        "nidn.training",
+        "nidn.training.losses",
+        "nidn.training.model",
+        "nidn.training.utils",
+        "nidn.trcwa",
+        "nidn.trcwa.utils",
+        "nidn.utils",
+    ],
+    python_requires=">=3.8, <4",
+    project_urls={
+        "Source": "https://github.com/esa/nidn/",
+    },
+)