removed the webuiGUI.py necessity, disable accelerate for windows (incompatible), remove some useless code (still learning)

.github/workflows/package.yml

@@ -34,5 +34,5 @@ jobs:
       - name: Upload artifacts
         uses: actions/upload-artifact@v3
         with:
-          name: ${{ matrix.os }}-binary-v1.5
+          name: ${{ matrix.os }}-binary-v1.5.1
           path: dist

StartUI.py

@@ -1,10 +1,10 @@
-import sys, os, gpustat, json, subprocess, platform, psutil, re, requests, darkdetect, qdarkstyle
+import sys, os, gpustat, json, subprocess, platform, psutil, re, requests, darkdetect, qdarkstyle, time
 from PyQt5.QtWidgets import QApplication, QHBoxLayout, QToolBar, QMessageBox, QAction, QMainWindow, QSpinBox, QLabel, QVBoxLayout, QComboBox, QSlider, QCheckBox, QLineEdit, QFileDialog, QPushButton, QWidget, QListWidget, QListWidgetItem, QGridLayout, QRadioButton, QFrame
 from PyQt5.QtCore import Qt
 from PyQt5.QtGui import QDoubleValidator, QIntValidator
 
 # For showing the current version and checking for updates
-version = "1.5"
+version = "1.5.1"
 
 # Profile folder for loading and saving profiles.
 profiles_folder = "./profiles"
@@ -15,6 +15,12 @@ model_folder = "./text-generation-webui/models"
 extensions_folder = "./text-generation-webui/extensions"
 loras_folder = "./text-generation-webui/loras"
 characters_folder = "./text-generation-webui/characters"
+
+if getattr(sys, 'frozen', False):
+    webui_file = sys._MEIPASS + '/webuiGUI.py'
+else:
+    webui_file = 'webuiGUI.py'
+
 # Get the current Max CPU threads to use, so the user can't exceed his thread count.
 max_threads = psutil.cpu_count(logical=True)
 
@@ -50,6 +56,7 @@ def run_cmd_with_conda(cmd, env=None):
 
         # Open a separate terminal window and execute the command
         subprocess.Popen(['start', 'cmd', '/k', full_cmd], shell=True, env=env)
+    
     elif platform.system() == 'Linux':
         # Define the necessary variables from the bash script
         install_dir = os.path.dirname(os.path.abspath(__file__))
@@ -90,7 +97,14 @@ class MainWindow(QMainWindow):
     def init_ui(self):
         self.setWindowTitle(f'StartUI for oobabooga webui v{version}')
 
-        # ToolBar
+        ##########################################
+        #  _____           _   ____              #
+        # |_   _|__   ___ | | | __ )  __ _ _ __  #
+        #   | |/ _ \ / _ \| | |  _ \ / _` | '__| #
+        #   | | (_) | (_) | | | |_) | (_| | |    #
+        #   |_|\___/ \___/|_| |____/ \__,_|_|    #
+        #                                        #
+        ##########################################
         toolbar = QToolBar()
         toolbar.setMovable(False)
         self.addToolBar(toolbar)
@@ -100,28 +114,31 @@ class MainWindow(QMainWindow):
         toolbar.addWidget(toolbar_label)
 
         # Deepspeed checkbox
-        self.deepspeed_settings_checkbox = QCheckBox("   DeepSpeed   ")
+        self.deepspeed_settings_checkbox = QCheckBox("\tDeepSpeed\t")
+        self.deepspeed_settings_checkbox.setToolTip("Enables specific DeepSpeed Settings.")
         self.deepspeed_settings_checkbox.setChecked(False)
-        self.deepspeed_settings_checkbox.setToolTip("Enables Deepspeed Settings")
         self.deepspeed_settings_checkbox.stateChanged.connect(self.on_deepspeed_settings_checkbox_stateChanged)
         toolbar.addWidget(self.deepspeed_settings_checkbox)
+        if platform.system() == 'Windows':
+            self.deepspeed_settings_checkbox.setEnabled(False)
+            self.deepspeed_settings_checkbox.setToolTip("DeepSpeed is not Supported in Windows.")
 
         # llama.cpp checkbox
-        self.llama_settings_checkbox = QCheckBox("   llama.cpp   ")
+        self.llama_settings_checkbox = QCheckBox("\tllama.cpp\t")
         self.llama_settings_checkbox.setChecked(False)
         self.llama_settings_checkbox.setToolTip("Enables llama.cpp Settings")
         self.llama_settings_checkbox.stateChanged.connect(self.on_llama_settings_checkbox_stateChanged)
         toolbar.addWidget(self.llama_settings_checkbox)
 
         # FlexGen Checkbox
-        self.flexgen_settings_checkbox = QCheckBox("   FlexGen   ")
+        self.flexgen_settings_checkbox = QCheckBox("\tFlexGen\t")
         self.flexgen_settings_checkbox.setChecked(False)
         self.flexgen_settings_checkbox.setToolTip("Enables FlexGen Settings")
         self.flexgen_settings_checkbox.stateChanged.connect(self.on_flexgen_settings_checkbox_stateChanged)
         toolbar.addWidget(self.flexgen_settings_checkbox)
 
         # RWKV Checkbox
-        self.rwkv_settings_checkbox = QCheckBox("   RWKV   ")
+        self.rwkv_settings_checkbox = QCheckBox("\tRWKV\t")
         self.rwkv_settings_checkbox.setChecked(False)
         self.rwkv_settings_checkbox.setVisible(False)
         self.rwkv_settings_checkbox.setToolTip("Enables RWKV Settings")
@@ -129,13 +146,30 @@ class MainWindow(QMainWindow):
         toolbar.addWidget(self.rwkv_settings_checkbox)
 
         # API Checkbox
-        self.api_settings_checkbox = QCheckBox("   API   ")
+        self.api_settings_checkbox = QCheckBox("\tAPI\t")
         self.api_settings_checkbox.setChecked(False)
         self.api_settings_checkbox.setToolTip("Enables API Settings")
         self.api_settings_checkbox.stateChanged.connect(self.on_api_settings_checkbox_stateChanged)
         toolbar.addWidget(self.api_settings_checkbox)
 
-        # Menu Bar
+        # Accelerate Checkbox
+        self.Accelerate_settings_checkbox = QCheckBox("\tAccelerate\t")
+        self.Accelerate_settings_checkbox.setChecked(False)
+        self.Accelerate_settings_checkbox.setToolTip("Enables API Settings")
+        self.Accelerate_settings_checkbox.stateChanged.connect(self.on_Accelerate_settings_checkbox_stateChanged)
+        toolbar.addWidget(self.Accelerate_settings_checkbox)
+        if platform.system() == 'Windows':
+            self.Accelerate_settings_checkbox.setEnabled(False)
+            self.Accelerate_settings_checkbox.setToolTip("Accelerate is not Supported in Windows.")
+
+        ################################################
+        #  __  __                    ____              #
+        # |  \/  | ___ _ __  _   _  | __ )  __ _ _ __  #
+        # | |\/| |/ _ \ '_ \| | | | |  _ \ / _` | '__| #
+        # | |  | |  __/ | | | |_| | | |_) | (_| | |    #
+        # |_|  |_|\___|_| |_|\__,_| |____/ \__,_|_|    #
+        #                                              #
+        ################################################
         menu = self.menuBar()
 
         # Main menu
@@ -179,7 +213,14 @@ class MainWindow(QMainWindow):
         report_bug_action.triggered.connect(self.on_report_bug_clicked)
         help_menu.addAction(report_bug_action)
 
-        # Main Window Layout, column width
+        ###################################################################
+        #  __  __       _        __        ___           _                #
+        # |  \/  | __ _(_)_ __   \ \      / (_)_ __   __| | _____      __ #
+        # | |\/| |/ _` | | '_ \   \ \ /\ / /| | '_ \ / _` |/ _ \ \ /\ / / #
+        # | |  | | (_| | | | | |   \ V  V / | | | | | (_| | (_) \ V  V /  #
+        # |_|  |_|\__,_|_|_| |_|    \_/\_/  |_|_| |_|\__,_|\___/ \_/\_/   #
+        #                                                                 #
+        ###################################################################
         layout = QGridLayout()
         layout.setColumnMinimumWidth(0, 350)
         layout.setColumnMinimumWidth(3, 30)
@@ -216,7 +257,6 @@ class MainWindow(QMainWindow):
         model_type_box.addWidget(self.model_type)
         layout.addLayout(model_type_box, 1, 0)
 
-
         # Character
         character_box = QHBoxLayout()
 
@@ -373,41 +413,26 @@ class MainWindow(QMainWindow):
         self.ram_slider.valueChanged.connect(self.on_ram_slider_changed)
 
         # Pre-layer Slider
-        # Check if Nvidia_gpu is enabled, if not, we don't need multiple pre_layer slider.
+        self.pre_layer_labels = []
-        if nvidia_gpu:
+        self.pre_layer_slider = []
-            self.pre_layer_labels = []
+        self.pre_layer_slider_value = []
-            self.pre_layer_slider = []
+        self.pre_layer_amount_max = 100
-            self.pre_layer_slider_value = []
+        # Don't get confused. With the latest changes, each GPU can have it's own pre_layer value. So we check again gpu_stats for the amount.
-            self.pre_layer_amount_max = 100
+        for i, gpu in enumerate(gpu_stats):
-            # Don't get confused. With the latest changes, each GPU can have it's own pre_layer value. So we check again gpu_stats for the amount.
+            pre_layer_labels = QLabel(f"{gpu.name} Pre_Layer:")
-            for i, gpu in enumerate(gpu_stats):
+            pre_layer_labels.setToolTip(f"The number of layers to allocate to the GPU.\nSetting this parameter enables CPU offloading for 4-bit models.\nFor multi-gpu, write the numbers separated by spaces, eg --pre_layer 30 60.")
-                pre_layer_labels = QLabel(f"{gpu.name} Pre_Layer:")
+            layout.addWidget(pre_layer_labels, 11 + (len(gpu_stats) * 2) + i, 0)
-                pre_layer_labels.setToolTip(f"The number of layers to allocate to the GPU.\nSetting this parameter enables CPU offloading for 4-bit models.\nFor multi-gpu, write the numbers separated by spaces, eg --pre_layer 30 60.")
+            self.pre_layer_labels.append(pre_layer_labels)
-                layout.addWidget(pre_layer_labels, 11 + (len(gpu_stats) * 2) + i, 0)
-                self.pre_layer_labels.append(pre_layer_labels)
 
-                pre_layer_sliders = QSlider(Qt.Horizontal)
+            pre_layer_sliders = QSlider(Qt.Horizontal)
-                pre_layer_sliders.setMaximum(100)
+            pre_layer_sliders.setMaximum(100)
-                pre_layer_sliders.valueChanged.connect(lambda value, idx=i: self.on_pre_layer_slider_changed(value, idx))
+            pre_layer_sliders.valueChanged.connect(lambda value, idx=i: self.on_pre_layer_slider_changed(value, idx))
-                layout.addWidget(pre_layer_sliders, 11 + (len(gpu_stats) * 2) + i, 1)
+            layout.addWidget(pre_layer_sliders, 11 + (len(gpu_stats) * 2) + i, 1)
-                self.pre_layer_slider.append(pre_layer_sliders)
+            self.pre_layer_slider.append(pre_layer_sliders)
 
-                pre_layer_sliders_value = QLabel("0")
+            pre_layer_sliders_value = QLabel("0")
-                layout.addWidget(pre_layer_sliders_value, 11 + (len(gpu_stats) * 2) + i, 2)
+            layout.addWidget(pre_layer_sliders_value, 11 + (len(gpu_stats) * 2) + i, 2)
-                self.pre_layer_slider_value.append(pre_layer_sliders_value)
+            self.pre_layer_slider_value.append(pre_layer_sliders_value)
-        else:
-            self.pre_layer_slider = QSlider(Qt.Horizontal)
-            self.pre_layer_slider.setMinimum(0)
-            self.pre_layer_slider.setMaximum(100)
-            self.pre_layer_slider.setTickInterval(1)
-            self.pre_layer_slider.setSingleStep(1)
-            layout.addWidget(QLabel("Pre-layer:"), 11 + len(gpu_stats), 0)
-            self.pre_layer_slider.setToolTip("The number of layers to allocate to the GPU. Setting this parameter enables CPU offloading for 4-bit models.")
-            layout.addWidget(self.pre_layer_slider, 11 + len(gpu_stats), 1)
-            self.pre_layer_slider.valueChanged.connect(self.on_pre_layer_slider_changed)
-
-            self.pre_layer_value_label = QLabel("0")
-            layout.addWidget(self.pre_layer_value_label, 11 + len(gpu_stats), 2)
 
         # Add horizontal line to seperate the Checkboxes
         line = QFrame()
@@ -522,7 +547,14 @@ class MainWindow(QMainWindow):
 
         # New GUI Options based on Toolbox Checkboxes.
 
-        # Deepspeed
+        ######################################################
+        #  ____                 ____                      _  #
+        # |  _ \  ___  ___ _ __/ ___| _ __   ___  ___  __| | #
+        # | | | |/ _ \/ _ \ '_ \___ \| '_ \ / _ \/ _ \/ _` | #
+        # | |_| |  __/  __/ |_) |__) | |_) |  __/  __/ (_| | #
+        # |____/ \___|\___| .__/____/| .__/ \___|\___|\__,_| #
+        #                 |_|        |_|                     #
+        ######################################################
 
         # Deepspeed Header
         self.deepspeed_label_header = QLabel("Deepspeed Options:")
@@ -601,7 +633,14 @@ class MainWindow(QMainWindow):
         self.deepspeed_line.setVisible(False)
         layout.addWidget(self.deepspeed_line, 36 + (len(gpu_stats) * 2), 0, 1, 3)
 
-        # llama.cpp
+        #################################################
+        #  _ _                                          #
+        # | | | __ _ _ __ ___   __ _   ___ _ __  _ __   #
+        # | | |/ _` | '_ ` _ \ / _` | / __| '_ \| '_ \  #
+        # | | | (_| | | | | | | (_| || (__| |_) | |_) | #
+        # |_|_|\__,_|_| |_| |_|\__,_(_)___| .__/| .__/  #
+        #                                 |_|   |_|     #
+        #################################################
 
         # llama.cpp Header
         self.llama_label_header = QLabel("llama.cpp Options:")
@@ -710,7 +749,14 @@ class MainWindow(QMainWindow):
         self.llama_line.setVisible(False)
         layout.addWidget(self.llama_line, 46 + (len(gpu_stats) * 2), 0, 1, 3)
 
-        # FlexGen Options
+        ########################################
+        #  _____ _            ____             #
+        # |  ___| | _____  __/ ___| ___ _ __   #
+        # | |_  | |/ _ \ \/ / |  _ / _ \ '_ \  #
+        # |  _| | |  __/>  <| |_| |  __/ | | | #
+        # |_|   |_|\___/_/\_\\____|\___|_| |_| #
+        #                                      #
+        ########################################
 
         # FlexGen Header Label
         self.flexgen_header_label = QLabel("FlexGen Options")
@@ -810,7 +856,14 @@ class MainWindow(QMainWindow):
         self.flexline.setVisible(False)
         layout.addWidget(self.flexline, 54 + (len(gpu_stats) * 2), 0, 1, 3)
 
-        # RWKV Options
+        ###################################
+        #  ______        ___  ____     __ #
+        # |  _ \ \      / / |/ /\ \   / / #
+        # | |_) \ \ /\ / /| ' /  \ \ / /  #
+        # |  _ < \ V  V / | . \   \ V /   #
+        # |_| \_\ \_/\_/  |_|\_\   \_/    #
+        #                                 #
+        ###################################
 
         # RWKV Header
         self.rwkv_header = QLabel("RWKV:")
@@ -865,7 +918,14 @@ class MainWindow(QMainWindow):
         self.rwkv_line.setVisible(False)
         layout.addWidget(self.rwkv_line, 65 + (len(gpu_stats) * 2), 0, 1, 3)
 
-        # API Options
+        ######################
+        #     _    ____ ___  #
+        #    / \  |  _ \_ _| #
+        #   / _ \ | |_) | |  #
+        #  / ___ \|  __/| |  #
+        # /_/   \_\_|  |___| #
+        #                    #
+        ######################
 
         # API Header Label
         self.api_header = QLabel("API:")
@@ -878,7 +938,6 @@ class MainWindow(QMainWindow):
         self.api_checkbox.setToolTip("Enable the API extension.")
         self.api_checkbox.setVisible(False)
         layout.addWidget(self.api_checkbox, 71 + (len(gpu_stats) * 2), 0)
-        #self.api_checkbox.stateChanged.connect(self.on_api_checkbox_changed)
 
         # API blocking Port Checkbox
         self.api_blocking_port_checkbox = QCheckBox("Change API Blocking Port")
@@ -919,12 +978,80 @@ class MainWindow(QMainWindow):
         layout.addWidget(self.api_public_checkbox, 74 + (len(gpu_stats) * 2), 0)
         self.api_public_checkbox.stateChanged.connect(self.on_api_public_checkbox_changed)
 
+        # Seperator for the Toolbox Options
+        self.toolboxapiline = QFrame()
+        self.toolboxapiline.setFrameShape(QFrame.HLine)
+        self.toolboxapiline.setFrameShadow(QFrame.Sunken)
+        self.toolboxapiline.setVisible(False)
+        layout.addWidget(self.toolboxapiline, 75 + (len(gpu_stats) * 2), 0, 1, 3)
+
+        #############################################################################
+        #     _                _                _         _  _         _     _ _    #
+        #    / \   ___ ___ ___| | ___ _ __ __ _| |_ ___  | || |       | |__ (_) |_  #
+        #   / _ \ / __/ __/ _ \ |/ _ \ '__/ _` | __/ _ \ | || |_ _____| '_ \| | __| #
+        #  / ___ \ (_| (_|  __/ |  __/ | | (_| | ||  __/ |__   _|_____| |_) | | |_  #
+        # /_/   \_\___\___\___|_|\___|_|  \__,_|\__\___|    |_|       |_.__/|_|\__| #
+        #                                                                           #
+        #############################################################################
+
+        # Accelerate 4-bit Header
+        self.accelerate4bit_header = QLabel("Accelerate 4-bit:")
+        self.accelerate4bit_header.setVisible(False)
+        self.accelerate4bit_header.setToolTip("Accelerate 4-bit: Choose the settings to use for accelerating 4-bit models.")
+        layout.addWidget(self.accelerate4bit_header, 80 + (len(gpu_stats) * 2), 0)
+
+        # Accelerate 4-bit Checkbox
+        self.accelerate4bit_checkbox = QCheckBox("Load in 4-bit")
+        self.accelerate4bit_checkbox.setToolTip("Load the model with 4-bit precision (using bitsandbytes).")
+        self.accelerate4bit_checkbox.setVisible(False)
+        layout.addWidget(self.accelerate4bit_checkbox, 81 + (len(gpu_stats) * 2), 0)
+
+        # Compute type horizontal layout
+        compute_type_layout = QHBoxLayout()
+
+        # Compute type label
+        self.accelerate4bit_compute_type_label = QLabel("Compute Type:")
+        self.accelerate4bit_compute_type_label.setToolTip("The compute type to use for 4-bit acceleration.")
+        self.accelerate4bit_compute_type_label.setVisible(False)
+        compute_type_layout.addWidget(self.accelerate4bit_compute_type_label)
+
+        # Compute type dropdown
+        self.accelerate4bit_compute_type_dropdown = QComboBox()
+        self.accelerate4bit_compute_type_dropdown.setToolTip("The compute type to use for 4-bit acceleration.")
+        self.accelerate4bit_compute_type_dropdown.setVisible(False)
+        self.accelerate4bit_compute_type_dropdown.addItems([ "none", "bfloat16", "float16", "float32"])
+        compute_type_layout.addWidget(self.accelerate4bit_compute_type_dropdown)
+        layout.addLayout(compute_type_layout, 81 + (len(gpu_stats) * 2), 1)
+
+        # Quant Type Horizontal Box
+        quant_type_layout = QHBoxLayout()
+
+        # Quant type label
+        self.accelerate4bit_quant_type_label = QLabel("Quant Type:")
+        self.accelerate4bit_quant_type_label.setToolTip("The quantization type to use for 4-bit acceleration.")
+        self.accelerate4bit_quant_type_label.setVisible(False)
+        quant_type_layout.addWidget(self.accelerate4bit_quant_type_label)
+
+        # Quant type Dropdown
+        self.accelerate4bit_quant_type_dropdown = QComboBox()
+        self.accelerate4bit_quant_type_dropdown.setToolTip("The quantization type to use for 4-bit acceleration.")
+        self.accelerate4bit_quant_type_dropdown.setVisible(False)
+        self.accelerate4bit_quant_type_dropdown.addItems([ "none", "nf4", "fp4"])
+        quant_type_layout.addWidget(self.accelerate4bit_quant_type_dropdown)
+        layout.addLayout(quant_type_layout, 82 + (len(gpu_stats) * 2), 1)
+
+        # Use double quant checkbox
+        self.accelerate4bit_double_quant_checkbox = QCheckBox("Use Double Quant")
+        self.accelerate4bit_double_quant_checkbox.setToolTip("Use double quantization for 4-bit acceleration.")
+        self.accelerate4bit_double_quant_checkbox.setVisible(False)
+        layout.addWidget(self.accelerate4bit_double_quant_checkbox, 82 + (len(gpu_stats) * 2), 0)
+
         # Seperator for the Toolbox Options
         self.toolboxendline = QFrame()
         self.toolboxendline.setFrameShape(QFrame.HLine)
         self.toolboxendline.setFrameShadow(QFrame.Sunken)
         self.toolboxendline.setVisible(False)
-        layout.addWidget(self.toolboxendline, 75 + (len(gpu_stats) * 2), 0, 1, 3)
+        layout.addWidget(self.toolboxendline, 84 + (len(gpu_stats) * 2), 0, 1, 3)
 
         # Authentication Box
         authentication_box = QHBoxLayout()
@@ -946,7 +1073,7 @@ class MainWindow(QMainWindow):
         self.choose_file_button.setToolTip("Choose a file to use for the authentication credentials. Credentials should be saved like:\nUSERNAME1:PASSWORD1\nUSERNAME2:PASSWORD2")
         self.choose_file_button.clicked.connect(self.on_choose_file_button_clicked)
         authentication_box.addWidget(self.choose_file_button)
-        layout.addLayout(authentication_box, 80 + (len(gpu_stats) * 2), 0, 1, 3)
+        layout.addLayout(authentication_box, 85 + (len(gpu_stats) * 2), 0, 1, 3)
 
         # Extensions Selection Menu
         self.use_extensions_checkbox = QCheckBox("Use Extensions")
@@ -1054,6 +1181,16 @@ class MainWindow(QMainWindow):
         central_widget.setLayout(layout)
         self.setCentralWidget(central_widget)
 
+    def on_Accelerate_settings_checkbox_stateChanged(self, state):
+        self.accelerate4bit_header.setVisible(state == Qt.Checked)
+        self.accelerate4bit_checkbox.setVisible(state == Qt.Checked)
+        self.accelerate4bit_compute_type_label.setVisible(state == Qt.Checked)
+        self.accelerate4bit_compute_type_dropdown.setVisible(state == Qt.Checked)
+        self.accelerate4bit_quant_type_label.setVisible(state == Qt.Checked)
+        self.accelerate4bit_quant_type_dropdown.setVisible(state == Qt.Checked)
+        self.accelerate4bit_double_quant_checkbox.setVisible(state == Qt.Checked)
+        self.toolboxendline.setVisible(state == Qt.Checked)
+
     def on_api_public_checkbox_changed(self, state):
         self.api_streaming_port_SpinBox.setEnabled(False)
         self.api_blocking_port_SpinBox.setEnabled(False)
@@ -1074,7 +1211,7 @@ class MainWindow(QMainWindow):
         self.api_streaming_port_checkbox.setVisible(state == Qt.Checked)
         self.api_streaming_port_SpinBox.setVisible(state == Qt.Checked)
         self.api_public_checkbox.setVisible(state == Qt.Checked)
-        self.toolboxendline.setVisible(state == Qt.Checked)
+        self.toolboxapiline.setVisible(state == Qt.Checked)
 
     def on_rwkv_settings_checkbox_stateChanged(self, state):
         self.rwkv_header.setVisible(state == Qt.Checked)
@@ -1099,8 +1236,6 @@ class MainWindow(QMainWindow):
         self.flexgen_pin_weight_label.setVisible(state == Qt.Checked)
         self.flexgen_pin_weight_dropdown.setVisible(state == Qt.Checked)
         self.flexline.setVisible(state == Qt.Checked)
-        #self.flexgen_line.setVisible(state == Qt.Checked)
-        #self.flexgen_line.setVisible(state == Qt.Checked)
 
     def on_llama_settings_checkbox_stateChanged(self, state):
         self.llama_label_header.setVisible(state == Qt.Checked)
@@ -1285,26 +1420,20 @@ class MainWindow(QMainWindow):
         self.ram_value_label.setText(f"{value} GiB")
 
     def on_pre_layer_slider_changed(self, value, idx):
-        if nvidia_gpu:
+        # Calculate the current total value of all sliders
-            # Calculate the current total value of all sliders
+        total_value = sum(slider.value() for slider in self.pre_layer_slider)
-            total_value = sum(slider.value() for slider in self.pre_layer_slider)
     
-            # Check if the total value exceeds the maximum
+        # Check if the total value exceeds the maximum
-            if total_value > self.pre_layer_amount_max:
+        if total_value > self.pre_layer_amount_max:
-                # Calculate the maximum allowed value for the current slider
+            # Calculate the maximum allowed value for the current slider
-                max_allowed_value = self.pre_layer_amount_max - (total_value - value)
+            max_allowed_value = self.pre_layer_amount_max - (total_value - value)
     
-                # Adjust the value of the current slider if necessary
+            # Adjust the value of the current slider if necessary
-                if value > max_allowed_value:
+            if value > max_allowed_value:
-                    self.pre_layer_slider[idx].setValue(max_allowed_value)
+                self.pre_layer_slider[idx].setValue(max_allowed_value)
-                    value = max_allowed_value
+                value = max_allowed_value
-        else:
-            # Update the value label with the current value of the pre-layer slider
-            self.pre_layer_value_label.setText(str(value))
     
-    def on_pre_layer_slider_changed(self, value):
+        self.pre_layer_slider_value[idx].setText(str(value))
-        # Update the value label with the current value of the pre-layer slider
-        self.pre_layer_value_label.setText(str(value))
 
     def on_vram_slider_changed(self, value, gpu_idx):
         self.gpu_vram_labels[gpu_idx].setText(f"{value} GiB")
@@ -1409,6 +1538,11 @@ class MainWindow(QMainWindow):
             "sdp_attention": self.use_sdp_attention_checkbox.isChecked(), # Saves the state of the sdp_attention checkbox
             "autogptq": self.use_autogptq_checkbox.isChecked(), # Saves the state of the autogptq checkbox
             "triton": self.use_triton_checkbox.isChecked(), # Saves the state of the triton checkbox
+            "acceleration": self.Accelerate_settings_checkbox.isChecked(), # Saves the state of the Accelerate checkbox
+            "use_4bit": self.accelerate4bit_checkbox.isChecked(), # Saves the state of the accelerate4bit checkbox
+            "compute_dtype": self.accelerate4bit_compute_type_dropdown.currentText(), # Saves the state of the accelerate4bit_compute_type_dropdown
+            "quant_type": self.accelerate4bit_quant_type_dropdown.currentText(), # Saves the state of the accelerate4bit_quant_type_dropdown
+            "use_x2_quant": self.accelerate4bit_double_quant_checkbox.isChecked(), # Saves the state of the accelerate4bit_double_quant_checkbox
             "deepspeed": self.deepspeed_settings_checkbox.isChecked(), # Saves the state of the deepspeed checkbox
             "deepspeed_enabled": self.deepspeed_checkbox.isChecked(), # Saves the state of the deepspeed checkbox
             "deepspeed_gpu_num": self.deepspeed_gpu_num_spinbox.value(), # Saves the state of the deepspeed_gpu_num_spinbox
@@ -1462,12 +1596,11 @@ class MainWindow(QMainWindow):
             "loras": [self.lora_list.item(i).text() for i in range(self.lora_list.count()) if self.lora_list.item(i).checkState() == Qt.Checked] # Saves the chosen loras
         }
 
+        pre_layer_values = [slider.value() for slider in self.pre_layer_slider]
+        settings["prelayer"] = pre_layer_values
+
         if nvidia_gpu:
             settings["gpu_vram"] = [slider.value() for slider in self.gpu_vram_sliders]
-            pre_layer_values = [slider.value() for slider in self.pre_layer_slider]
-            settings["prelayer"] = pre_layer_values
-        else:
-            settings["prelayer"] = self.pre_layer_value_label.text()
 
         # Get the text entered in the text field
         profile_name = self.profile_name_textfield.text()
@@ -1479,7 +1612,6 @@ class MainWindow(QMainWindow):
 
     def expression_check(self, command):
         selected_model = self.model_dropdown.currentText()
-        #print(f"Selected model: {selected_model}")
         
         # Use a regular expression to check if the selected model matches the pattern
         if re.search(r".*mpt.*7b", selected_model, re.IGNORECASE):
@@ -1525,9 +1657,8 @@ class MainWindow(QMainWindow):
             command += f" --model {chosen_model}"
 
         # Add the chosen model type to the command
-        chosen_model_type = self.model_type.currentText()
         if self.model_type.currentText() != "none" and self.model_dropdown.currentText() != "none":
-            command += f" --model_type {chosen_model_type}"
+            command += f" --model_type {self.model_type.currentText()}"
 
         # Add loras to the command
         loras = [self.lora_list.item(i).text() for i in range(self.lora_list.count()) if self.lora_list.item(i).checkState() == Qt.Checked]
@@ -1536,22 +1667,18 @@ class MainWindow(QMainWindow):
                 command += f" --lora {' '.join(loras)}"
 
         # Add Characters to the command
-        chosen_characters = self.character_to_load.currentText()
         if self.character_to_load.currentText() != "none":
-            command += f" --character {chosen_characters}"
+            command += f" --character {self.character_to_load.currentText()}"
-            print(chosen_characters)
 
         # Adds wbits to the command, if not "none"
-        chosen_wbits = self.wbit_dropdown.currentText()
         if self.wbit_dropdown.currentText() != "none":
             if not self.cpu_radio_button.isChecked() and self.model_dropdown.currentText() != "none":
-                command += f" --wbits {chosen_wbits}"
+                command += f" --wbits {self.wbit_dropdown.currentText()}"
 
         # Adds Groupsize to the command, if not "none"
-        chosen_gsize = self.gsize_dropdown.currentText()
         if self.gsize_dropdown.currentText() != "none":
             if not self.cpu_radio_button.isChecked() and self.model_dropdown.currentText() != "none":
-                command += f" --groupsize {chosen_gsize}"
+                command += f" --groupsize {self.gsize_dropdown.currentText()}"
 
         # Add the chosen mode to the command (Chat, cai-chat, notebook)
         chosen_mode = self.mode_dropdown.currentText()
@@ -1620,6 +1747,21 @@ class MainWindow(QMainWindow):
         if self.use_quant_checkbox.isChecked():
             command += " --quant_attn"
 
+        # Accelerate 4-bit
+
+        # 4-bit usage
+        if self.accelerate4bit_checkbox.isChecked():
+            command += " --load-in-4bit"
+
+        if self.accelerate4bit_compute_type_dropdown.currentText() != "none":
+            command += f" --compute_dtype {self.accelerate4bit_compute_type_dropdown.currentText()}"
+
+        if self.accelerate4bit_quant_type_dropdown.currentText() != "none":
+            command += f" --quant_type {self.accelerate4bit_quant_type_dropdown.currentText()}"
+
+        if self.accelerate4bit_double_quant_checkbox.isChecked():
+            command += " --use_double_quant"
+
         # Disable Cache
         if self.use_nocache_checkbox.isChecked():
             command += " --no-cache"
@@ -1655,13 +1797,9 @@ class MainWindow(QMainWindow):
                 command += f" --gradio-auth-path {self.choose_file_label.text()}"
 
         ## Adds the Prelayer selection
-        if nvidia_gpu:
+        slider_values = [slider.value() for slider in self.pre_layer_slider]
-            slider_values = [slider.value() for slider in self.pre_layer_slider]
+        if any(value > 0 for value in slider_values):
-            if any(value > 0 for value in slider_values):
+            command += f" --pre_layer {' '.join(str(value) for value in slider_values if value > 0)}"
-                command += f" --pre_layer {' '.join(str(value) for value in slider_values if value > 0)}"
-        else:
-            if int(self.pre_layer_value_label.text()) > 0:
-                command += f" --pre_layer {self.pre_layer_value_label.text()}"
 
         # IF sdp_attention is checked
         if self.use_sdp_attention_checkbox.isChecked():
@@ -1692,7 +1830,7 @@ class MainWindow(QMainWindow):
                 command += f" --api-streaming-port {self.api_streaming_port_SpinBox.text()}"
 
         # Just for debugging.
-        #print(f"Command generated: python webuiGUI.py {command}")
+        print(f"Command generated: python {webui_file} {command}")
 
         # Based on the Model that's chosen, we will take care of some necessary stuff.
         # Starts the webui in the conda env with the user given Options
@@ -1711,13 +1849,16 @@ class MainWindow(QMainWindow):
                 QMessageBox.critical(self, "Error", message)
 
         if not self.deepspeed_checkbox.isChecked():
-            run_cmd_with_conda(f"python webuiGUI.py {command}")
+            if self.use_8bit_checkbox.isChecked():
+                run_cmd_with_conda(f"pip install accelerate && python {webui_file} {command}")
+            else:
+                run_cmd_with_conda(f"python {webui_file} {command}")
 
         if self.use_autoclose_checkbox.isChecked():
             sys.exit()
             
     def on_update_button_clicked(self):
-        run_cmd_with_conda("python webuiGUI.py --update && exit")
+        run_cmd_with_conda(f"python {webui_file} --update && exit")
 
     def load_profile(self, profile_file):
         with open(profile_file, "r") as file:
@@ -1770,6 +1911,11 @@ class MainWindow(QMainWindow):
         self.use_sdp_attention_checkbox.setChecked(settings.get("sdp_attention", False))
         self.use_autogptq_checkbox.setChecked(settings.get("autogptq", False))
         self.use_triton_checkbox.setChecked(settings.get("triton", False))
+        self.Accelerate_settings_checkbox.setChecked(settings.get("acceleration", False))
+        self.accelerate4bit_checkbox.setChecked(settings.get("use_4bit", False))
+        self.accelerate4bit_compute_type_dropdown.setCurrentText(settings.get("compute_dtype", ""))
+        self.accelerate4bit_quant_type_dropdown.setCurrentText(settings.get("quant_type", ""))
+        self.accelerate4bit_double_quant_checkbox.setChecked(settings.get("use_x2_quant", False))
         self.deepspeed_settings_checkbox.setChecked(settings.get("deepspeed", False))
         self.deepspeed_checkbox.setChecked(settings.get("deepspeed_enabled", False))
         self.deepspeed_gpu_num_spinbox.setValue(int(settings.get("deepspeed_gpu_num", 0)))
@@ -1815,16 +1961,14 @@ class MainWindow(QMainWindow):
         self.authentication_checkbox.setChecked(settings.get("authentication", False))
         self.choose_file_label.setText(settings.get("authentication_file", ""))
         self.character_to_load.setCurrentText(settings.get("character", ""))
+        #self.pre_layer_slider.setValue(int(settings.get("prelayer", 0)))
         self.use_autolaunch_checkbox.setChecked(settings.get("autolaunch", False))
         self.use_network_checkbox.setChecked(settings.get("listen", False))
 
-        if nvidia_gpu:
+        if "prelayer" in settings:
-            if "prelayer" in settings:
+            pre_layer_values = settings["prelayer"]
-                pre_layer_values = settings["prelayer"]
+            for i, value in enumerate(pre_layer_values):
-                for i, value in enumerate(pre_layer_values):
+                self.pre_layer_slider[i].setValue(value)
-                    self.pre_layer_slider[i].setValue(value)
-        else:
-            self.pre_layer_slider.setValue(int(settings.get("prelayer", 0)))
 
         if nvidia_gpu:
             gpu_vram_settings = settings.get("gpu_vram", [])