How to run the code on multi gpus?

[mvoofan]

This project is a fantastic work on non-surpervised clustring.

However, OOM (out of memory) errors appeared when I attempted to run the codes with my own data (feather size per sample > 20000). Then I tried to solve the problem by running the codes on multi gpus, and added the following data in 'Conv.py'.

    self.train_step1_keras=tf.keras.optimizers.Adam()
    self.train_step1_keras.minimize(  self.loss_defense,var_list=[self.vae.weights,self.Discriminator.weights] )

def fit_defense(self, x_train_unlabeled,x_dy,x_dy1,batch_size):
    defense_model_parrarel=multi_gpu_model(self.defense_model, gpus = self.gpu)
    defense_model_parrarel.compile(optimizer=self.train_step1_keras)
    defense_model_parrarel.fit(x=[ x_train_unlabeled,x_dy,x_dy1], batch_size=batch_size, steps_per_epoch=100 )

Then " returned a result with an error set" error appeared when calling "self.train_step1_keras.minimize( self.loss_defense,var_list=[self.vae.weights,self.Discriminator.weights] )".

I wonder if anyone can provide any suggestions on running the code on multi gpus.

[LYK-love]

    self.train_step1_keras=tf.keras.optimizers.Adam()
    self.train_step1_keras.minimize(  self.loss_defense,var_list=[self.vae.weights,self.Discriminator.weights] )

def fit_defense(self, x_train_unlabeled,x_dy,x_dy1,batch_size):
    defense_model_parrarel=multi_gpu_model(self.defense_model, gpus = self.gpu)
    defense_model_parrarel.compile(optimizer=self.train_step1_keras)
    defense_model_parrarel.fit(x=[ x_train_unlabeled,x_dy,x_dy1], batch_size=batch_size, steps_per_epoch=100 )

Then " returned a result with an error set" error appeared when calling "self.train_step1_keras.minimize( self.loss_defense,var_list=[self.vae.weights,self.Discriminator.weights] )".

I wonder if anyone can provide any suggestions on running the

Hi bro, i also fork this project today. However, i found the sentence "import cv2 " in **/src/applications/ADDC.py. And i can't find this module online or in the files downloaded. I strongly doubt that the "cv2" is written locally and was never uploaded., which means the code can never be runned by just typing "python run.py ". However, i saw you made this issue which inpliied you just run the code successfully on the original dataset. I wonder that how do you make it? Did you write your own cv2 or its just something wrong with me?

[mvoofan]

    self.train_step1_keras=tf.keras.optimizers.Adam()
    self.train_step1_keras.minimize(  self.loss_defense,var_list=[self.vae.weights,self.Discriminator.weights] )

def fit_defense(self, x_train_unlabeled,x_dy,x_dy1,batch_size):
    defense_model_parrarel=multi_gpu_model(self.defense_model, gpus = self.gpu)
    defense_model_parrarel.compile(optimizer=self.train_step1_keras)
    defense_model_parrarel.fit(x=[ x_train_unlabeled,x_dy,x_dy1], batch_size=batch_size, steps_per_epoch=100 )

Then " returned a result with an error set" error appeared when calling "self.train_step1_keras.minimize( self.loss_defense,var_list=[self.vae.weights,self.Discriminator.weights] )".
I wonder if anyone can provide any suggestions on running the

Hi bro, i also fork this project today. However, i found the sentence "import cv2 " in **/src/applications/ADDC.py. And i can't find this module online or in the files downloaded. I strongly doubt that the "cv2" is written locally and was never uploaded., which means the code can never be runned by just typing "python run.py ". However, i saw you made this issue which inpliied you just run the code successfully on the original dataset. I wonder that how do you make it? Did you write your own cv2 or its just something wrong with me?

it is the "opencv-python" package~