search

yousifj129 / NeuralNetworkCreator

Create Sequential Neural Networks Using a GUI program
GNU General Public License v3.0
2 stars 1 forks source link

the user can only choose one activation function for all layers #1

Open yousifj129 opened 2 months ago

yousifj129 commented 2 months ago

the user can only choose one activation function for all layers, make the user able to choose different activation function whenever he wants

ThePythonDev commented 1 month ago

What do you mean? I can try to help with this issue as I am very interested in this repository and its function

yousifj129 commented 1 month ago

What do you mean? I can try to help with this issue as I am very interested in this repository and its function

as you see, we have activation functions we can use: self.activation = QComboBox() self.activation.addItems(["relu", "sigmoid", "softmax", "tanh", "linear", "softplus","softsign","selu","elu","exponential","leaky_relu", "relu6","silu","hard_silu","gelu","hard_sigmoid","mish","log_softmax"]) layer_layout.addWidget(QLabel("Activation Function:")) layer_layout.addWidget(self.activation)

and in the create model function: def create_model(self): self.model = keras.Sequential() self.model.add(keras.layers.Input(shape=(self.input_dim,))) activ = self.activation.currentText() for i in range(self.layer_list.count()): layer_str = self.layer_list.item(i).text() layer_type, params = layer_str.split(': ')

        if layer_type == "Dense":
            units = int(params)
            self.model.add(keras.layers.Dense(units, activation=activ))
        elif layer_type == "Conv1D":
            filters, kernel_size = map(int, params.split(','))
            self.model.add(keras.layers.Conv1D(filters, kernel_size, activation=activ))
        elif layer_type == "LSTM":
            units = int(params)
            self.model.add(keras.layers.LSTM(units, activation=activ))
        elif layer_type == "Dropout":
            rate = float(params)
            self.model.add(keras.layers.Dropout(rate))

    self.model.add(keras.layers.Dense(self.output_dim))

   we are taking the current activ function, but we take it for every layer, which is a problem, we should make some list variable that contains each layer activation function so each layer can have its own activation
ThePythonDev commented 1 month ago

I will test a few things and see if I can try to help with this.

EDIT: Is this built for Windows? If so I am sorry but I run linux

yousifj129 commented 1 month ago

I will test a few things and see if I can try to help with this.

EDIT: Is this built for Windows? If so I am sorry but I run linux

I think it should run just fine on Linux, it's just tensorflow and pyside6