pmap` support on cpu with following code: `os.environ['XLA_FLAGS'] = '--xla_force_host_platform_device_count=24'`. fix examples/ucell_optimize.py
0.7.3
added fft_type in args of RCWA Class; eig perturbation code correction in JAX and Torch; output of self.solve changed because of JAX jit compilation.
overall cleaning; updating examples;
0.7.2
update ex_ucell.py
add fft_type in run_ucell
update readme
add fft_method_matching.py in benchmarks
update ucell_optimize
enforce fourier_order to be an integer type
0.7.1
* Added Automatic Differentiation in JAX * Separated eig64 and eig32 in JAX. * Rewriting and cleaning conv_mat * Updated conv_mat method * Minimized minimum_pattern_size * Updating examples/optimization/ucell_optimize.py * Converted ucell dtype explicitly * Reorganized code * Added '.gitattributes' to hide ipynb files from github stats.
0.7.0
Optimized JAX code (jit was applied in broad range). Added JAX vmap and pmap. Example code for TPU backend development(on hold due to 'NotImplementedError') Added Citation file host_callback was replaced by pure_callback in eig in JAX due to slow speed in pmap
0.6.0
Supports Numpy, JAX and PyTorch as a backend for EM simulation and its optimization.