Dev

.gitignore

@@ -1,4 +1,8 @@
 .idea
 *.egg-info
-Miniconda3-py310_22.11.1-1-Linux-x86_64.sh
-.vscode/launch.json
+.vscode
+.DS_Store
+
+venv/
+build/
+dist/

QA/backprop.py

@@ -11,7 +11,7 @@
 import time
 from copy import deepcopy
 
-from meent.entrance import call_solver
+from meent.main import call_solver
 import torch
 
 

QA/run_debug.py

@@ -1,7 +1,7 @@
 import time
 import numpy as np
 
-from meent.entrance import RCWA
+from meent.main import RCWA
 
 
 def run_debug_cases(n_I, n_II, theta, phi, grating_type, pol):

QA/run_test.py

@@ -1,7 +1,7 @@
 import time
 import numpy as np
 
-from meent.entrance import call_solver
+from meent.main import call_solver
 
 
 def run_test(n_I, n_II, theta, phi, grating_type, pol):

QA/run_test2.py

@@ -1,6 +1,6 @@
 import numpy as np
 
-from meent.entrance import call_solver, sweep_wavelength
+from meent.main import call_solver, sweep_wavelength
 
 
 pol = 1  # 0: TE, 1: TM

QA/test_case.py

@@ -1,6 +1,6 @@
 import numpy as np
 
-from meent.entrance import call_solver
+from meent.main import call_solver
 
 
 def test():

README.md

@@ -5,26 +5,19 @@
 Meent is a RCWA solver and its applications on optimization problem. We are expecting that this tool can accelerate ML research in photonics.
 
 ## How to install
-
-You can install from PyPI
-
 ```shell
 pip install meent
 ```
 
-or download this repo and run
-
-```shell
-pip install .
-```
+JAX and PyTorch is needed for advanced utilization.
 
 ## How to use
 
 Meent provides Numpy, JAX and PyTorch as a backend.
 
 ```python
 import meent
-from meent.entrance import call_solver
+from meent.main import call_solver
 
 # mode 0 = Numpy
 # mode 1 = JAX
@@ -34,6 +27,7 @@ mode_key = 1
 solver = meent.rcwa.call_solver(mode=mode_key, ...)
 ```
 
+
 ## When to use
 
 |                 | Numpy | JAX | PyTorch |

benchmarks/README.md

@@ -1,3 +1,12 @@
 # Benchmark
-run EM simulation in Python process.
+Compare Meent to other EM simulation SW. \
+This is only to provide technical report and discussion, not to say any tool is good or bad.
+
+## 1. Fourier transform
+convergence test with both of Discrete FT and Continuous FT.
+
+## 2. Calculation time
+time to get diffraction efficiencies from EM solver.
+
+## 3. Eigendecomposition with CPU core limit
 

benchmarks/fft_method_comparison.py

@@ -10,7 +10,7 @@
 import jax.numpy as jnp
 import time
 
-from meent.entrance import call_solver
+from meent.main import call_solver
 import torch
 
 
@@ -116,8 +116,8 @@ def compare_conv_mat_method(mode_key, dtype, device):
         from meent.on_numpy.emsolver.convolution_matrix import to_conv_mat_continuous as conv2
 
     elif mode_key == 1:
-        from meent.on_jax.convolution_matrix import to_conv_mat_discrete as conv1
-        from meent.on_jax.convolution_matrix import to_conv_mat_continuous as conv2
+        from meent.on_jax.emsolver.convolution_matrix import to_conv_mat_discrete as conv1
+        from meent.on_jax.emsolver.convolution_matrix import to_conv_mat_continuous as conv2
     else:
         from meent.on_torch.emsolver.convolution_matrix import to_conv_mat_discrete as conv1
         from meent.on_torch.emsolver.convolution_matrix import to_conv_mat_continuous as conv2