complex.py

import os
os.environ["CUDA_VISIBLE_DEVICES"] = ""

import nobuco
from nobuco.layers.weight import WeightLayer

import torch
from torch import nn

import tensorflow as tf
from tensorflow.lite.python.lite import TFLiteConverter
import keras


class ModelComplex(nn.Module):

    def __init__(self):
        super().__init__()
        self.mask = torch.randn(1, 3, 100, 100).to(torch.complex64)

    def forward(self, x):
        x = torch.complex(x, x)
        x = x.to(torch.complex128)
        x = x * self.mask

        s1 = torch.std(x, dim=1, unbiased=False, keepdim=True)
        s2 = x.std(dim=3, unbiased=True, keepdim=False)

        x = x.view(1, -1)
        x = x.t()
        return x, s1, s2


model = ModelComplex()

dummy_image = torch.randn(1, 3, 100, 100)

keras_model = nobuco.pytorch_to_keras(
    model,
    args=[dummy_image],
)

model_path = 'complex'
keras_model.save(model_path + '.h5')
print('Model saved')

custom_objects = {'WeightLayer': WeightLayer}

keras_model_restored = keras.models.load_model(model_path + '.h5', custom_objects=custom_objects)
print('Model loaded')

converter = TFLiteConverter.from_keras_model_file(model_path + '.h5', custom_objects=custom_objects)
converter.target_ops = [tf.lite.OpsSet.SELECT_TF_OPS, tf.lite.OpsSet.TFLITE_BUILTINS]
tflite_model = converter.convert()
with open(model_path + '.tflite', 'wb') as f:
    f.write(tflite_model)