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 = "1.6"
# Profile folder for loading and saving profiles.
profiles_folder = "./profiles"
# Create the profile folder if it doesn't exist
os.makedirs(profiles_folder, exist_ok=True)
repo_path = "./text-generation-webui"
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)
# Check if Nvidia GPU and driver is installed, set boolean for later references
try:
output = subprocess.check_output(['nvidia-smi'])
nvidia_gpu = True
except:
nvidia_gpu = False
pass
# # Get the absolute path of the script file
script_path = os.path.abspath(__file__)
# Define the path of the settings file relative to the script file
settings_file = os.path.join(os.path.dirname(script_path), "gui-config.json")
# Define the conda environment path
if platform.system() == 'Windows':
# Sets the Conda Environment based on Windows
conda_binary = r".\installer_files\conda\condabin\conda.bat"
conda_env_path = r".\installer_files\env"
if platform.system() == 'Linux':
# Sets the Conda Environment based on Linux
conda_binary = "./installer_files/conda/condabin/conda"
conda_env_path = "./installer_files/env"
def run_cmd_with_conda(cmd, env=None):
if platform.system() == 'Windows':
# For Windows, activate the Conda environment using the activate.bat script
activate_cmd = f"{conda_binary} activate {conda_env_path} && "
full_cmd = activate_cmd + cmd
# 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__))
conda_root_prefix = os.path.join(install_dir, "installer_files", "conda")
install_env_dir = os.path.join(install_dir, "installer_files", "env")
# For Linux, activate the Conda environment
activate_cmd = f"source {os.path.join(conda_root_prefix, 'etc', 'profile.d', 'conda.sh')} && conda activate {install_env_dir}"
# Check for available terminal emulators
terminal_emulators = ['xdg-terminal', 'gnome-terminal', 'konsole', 'xfce4-terminal', 'mate-terminal', 'lxterminal', 'termite', 'tilix', 'xterm']
terminal_cmd = None
for emulator in terminal_emulators:
try:
subprocess.run([emulator, '--version'], check=True)
terminal_cmd = emulator
break
except FileNotFoundError:
continue
if terminal_cmd is None:
raise RuntimeError("No compatible terminal emulator found.")
# Execute the command within the Conda environment in a separate terminal
print(cmd)
subprocess.Popen([terminal_cmd, '--', 'bash', '-c', f"{activate_cmd} && {cmd}"], env=env)
class MainWindow(QMainWindow):
def __init__(self):
super().__init__()
self.init_ui()
self.load_settings()
self.set_ram_slider_max()
self.update_check()
def init_ui(self):
self.setWindowTitle(f'StartUI for oobabooga webui v{version}')
##########################################
# _____ _ ____ #
# |_ _|__ ___ | | | __ ) __ _ _ __ #
# | |/ _ \ / _ \| | | _ \ / _` | '__| #
# | | (_) | (_) | | | |_) | (_| | | #
# |_|\___/ \___/|_| |____/ \__,_|_| #
# #
##########################################
toolbar = QToolBar()
toolbar.setMovable(False)
self.addToolBar(toolbar)
# Toolbar Label
toolbar_label = QLabel("Show Advanced Settings:")
toolbar.addWidget(toolbar_label)
# Deepspeed checkbox
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.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("\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("\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("\tRWKV\t")
self.rwkv_settings_checkbox.setChecked(False)
self.rwkv_settings_checkbox.setVisible(False)
self.rwkv_settings_checkbox.setToolTip("Enables RWKV Settings")
self.rwkv_settings_checkbox.stateChanged.connect(self.on_rwkv_settings_checkbox_stateChanged)
toolbar.addWidget(self.rwkv_settings_checkbox)
# API Checkbox
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)
# 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
main_menu = menu.addMenu("StartUI")
main_menu.addAction("Exit", self.close)
# help menu
help_menu = menu.addMenu("Help")
# Help menu actions
# Github action
github_action = QAction("Github", self)
github_action.setStatusTip("Opens the Github Page")
github_action.triggered.connect(self.on_Github_clicked)
help_menu.addAction(github_action)
# Oobabooga action
oobabooga_action = QAction("oobabooga", self)
oobabooga_action.setStatusTip("Opens the oobabooga Github Page")
oobabooga_action.triggered.connect(self.on_oobabooga_clicked)
help_menu.addAction(oobabooga_action)
# Version action
version_action = QAction(f"Version: {version}", self)
version_action.setStatusTip("Shows the Version of StartUI")
help_menu.addAction(version_action)
version_action.triggered.connect(self.show_version_window)
# About Action
about_action = QAction("About", self)
about_action.setToolTip("Opens the About Page")
about_action.triggered.connect(self.show_about_window)
help_menu.addAction(about_action)
# seperator
help_menu.addSeparator()
# Report Bug
report_bug_action = QAction("Report Bug", self)
report_bug_action.setToolTip("Opens the Github Issue Page with creating a new issue")
report_bug_action.triggered.connect(self.on_report_bug_clicked)
help_menu.addAction(report_bug_action)
###################################################################
# __ __ _ __ ___ _ #
# | \/ | __ _(_)_ __ \ \ / (_)_ __ __| | _____ __ #
# | |\/| |/ _` | | '_ \ \ \ /\ / /| | '_ \ / _` |/ _ \ \ /\ / / #
# | | | | (_| | | | | | \ V V / | | | | | (_| | (_) \ V V / #
# |_| |_|\__,_|_|_| |_| \_/\_/ |_|_| |_|\__,_|\___/ \_/\_/ #
# #
###################################################################
layout = QGridLayout()
layout.setColumnMinimumWidth(0, 350)
layout.setColumnMinimumWidth(3, 30)
# Model Dropdown
# Get the list of models in models folder
model_box = QHBoxLayout()
model_folders = [name for name in os.listdir(model_folder) if os.path.isdir(os.path.join(model_folder, name))]
self.model_dropdown = QComboBox()
self.model_dropdown.addItem("none")
self.model_dropdown.addItems(model_folders)
self.model_dropdown.setToolTip("Select your prefered Model")
model_box.addWidget(QLabel("Choose Model:"))
model_box.addWidget(self.model_dropdown)
layout.addLayout(model_box, 0, 0)
# Reload Model Button
self.reload_model_button = QPushButton("Reload the Model List")
self.reload_model_button.setToolTip("Reloads the Names in the Models Folder")
self.reload_model_button.clicked.connect(self.reload_models)
layout.addWidget(self.reload_model_button, 0, 1, 1, 2)
# Model Type
model_type_box = QHBoxLayout()
# Model Type Label
self.model_type_text = QLabel("Model Type:")
model_type_box.addWidget(self.model_type_text)
# Model Type Dropdown
self.model_type = QComboBox()
self.model_type.addItems(["none", "llama", "opt", "gptj"])
self.model_type.setToolTip("Select the Model Type")
model_type_box.addWidget(self.model_type)
layout.addLayout(model_type_box, 1, 0)
# Character
character_box = QHBoxLayout()
# Character Text
self.character_text = QLabel("Character:")
character_box.addWidget(self.character_text)
# Character Dropdown
self.character_to_load = QComboBox()
# get a list of all .json files in the characters folder
character_jsons = [file for file in os.listdir(characters_folder) if file.endswith(".json")]
without_suffix = [file.replace(".json", "") for file in character_jsons]
self.character_to_load.addItem("none")
self.character_to_load.addItems(without_suffix)
self.character_to_load.setToolTip("Select the Character you want to load")
character_box.addWidget(self.character_to_load)
layout.addLayout(character_box, 1, 1, 1, 2)
# WBIT Box
wbit_box = QHBoxLayout()
# WBIT Label
self.wbit_text = QLabel("Choose WBITs:")
wbit_box.addWidget(self.wbit_text)
# WBIT Dropdown Menu
self.wbit_dropdown = QComboBox()
self.wbit_dropdown.addItems(["none", "1", "2", "3", "4","8"])
self.wbit_dropdown.setToolTip("Select the bits quantization for this model\nExample: vicuna 7b 4bit you should choose 4.\nYou can keep it at none, the webui will determine it automatically if the wbits are mentioned in the name of the model")
wbit_box.addWidget(self.wbit_dropdown)
layout.addLayout(wbit_box, 2, 0)
# Groupsize box
groupsize_box = QHBoxLayout()
# Groupsize Label
self.groupsize_text = QLabel("Choose Groupsize:")
groupsize_box.addWidget(self.groupsize_text)
# Groupsize Dropdown Menu
self.gsize_dropdown = QComboBox()
self.gsize_dropdown.addItems(["none", "32", "64", "128", "1024"])
self.gsize_dropdown.setToolTip("Select the groupsize used by the Model.\nExample: vicuna 7b 4bit-128g you should choose 128.\nYou can keep it at none, the webui will determine it automatically if the groupsize is mentioned in the name of the model")
groupsize_box.addWidget(self.gsize_dropdown)
layout.addLayout(groupsize_box, 2, 1, 1, 2)
# Interface Mode Box
interface_mode_box = QHBoxLayout()
# Interface mode label
self.interface_mode_text = QLabel("Interface Mode:")
interface_mode_box.addWidget(self.interface_mode_text)
# Interface Mode Dropdown
self.mode_dropdown = QComboBox()
self.mode_dropdown.addItems(["chat", "notebook"])
self.mode_dropdown.setToolTip("Choose what kind of Interface you want to load.")
interface_mode_box.addWidget(self.mode_dropdown)
layout.addLayout(interface_mode_box, 3, 0)
# WebUI Update
self.update_button = QPushButton("Update the text-generation-webui")
self.update_button.setToolTip("Starts the Update Routine for the text-generation-webui")
self.update_button.clicked.connect(self.on_update_button_clicked)
layout.addWidget(self.update_button, 3, 1, 1, 2)
# Add horizontal line to seperate the CPU/GPU Settings
line = QFrame()
line.setFrameShape(QFrame.HLine)
line.setFrameShadow(QFrame.Sunken)
layout.addWidget(line, 9, 0, 1, 3)
# GPU Checkbox and Sliders
self.gpu_radio_button = QRadioButton("Use GPU")
if nvidia_gpu:
# If nvidia_gpu is true, enable the gpu radio button
self.gpu_radio_button.setChecked(True)
self.gpu_radio_button.setToolTip("Choose if you want to use your GPU")
else:
# If nvidia_gpu is false, disable the gpu radio button
self.gpu_radio_button.setToolTip("AMD or Intel GPU's are currently not supported.")
self.gpu_radio_button.setChecked(False)
self.gpu_radio_button.setEnabled(False)
layout.addWidget(self.gpu_radio_button, 10, 0)
self.cpu_radio_button = QRadioButton("Use CPU")
self.cpu_radio_button.setToolTip("Choose if you want to use your CPU")
self.cpu_radio_button.setChecked(False)
layout.addWidget(self.cpu_radio_button, 10, 1)
self.auto_radio_button = QRadioButton("Autodevice")
self.auto_radio_button.setToolTip("Let the webui decide whats best for you!")
if nvidia_gpu:
self.auto_radio_button.setChecked(False)
else:
self.auto_radio_button.setChecked(True)
layout.addWidget(self.auto_radio_button, 10, 2)
# Connect the radio button to functions
self.gpu_radio_button.toggled.connect(self.on_gpu_radio_button_toggled)
self.cpu_radio_button.toggled.connect(self.on_cpu_radio_button_toggled)
self.auto_radio_button.toggled.connect(self.on_auto_radio_button_toggled)
# Get GPU Information and sliders for nvidia_gpus
if nvidia_gpu:
self.gpu_vram_sliders = []
self.gpu_vram_labels = []
self.gpu_labels = []
gpu_stats = gpustat.GPUStatCollection.new_query()
for i, gpu in enumerate(gpu_stats):
gpu_label = QLabel(f"{gpu.name} VRAM:")
gpu_label.setToolTip(f"Total VRAM: {gpu.memory_total} MiB\nUsed VRAM: {gpu.memory_used} MiB\nFree VRAM: {gpu.memory_free} MiB")
layout.addWidget(gpu_label, 11 + i, 0)
self.gpu_labels.append(gpu_label)
vram_slider = QSlider(Qt.Horizontal)
vram_slider.setMaximum(int(gpu.memory_total / 1024))
vram_slider.valueChanged.connect(lambda value, idx=i: self.on_vram_slider_changed(value, idx))
layout.addWidget(vram_slider, 11 + i, 1)
vram_value_label = QLabel("0 GiB")
layout.addWidget(vram_value_label, 11 + i, 2)
self.gpu_vram_labels.append(vram_value_label)
self.gpu_vram_sliders.append(vram_slider)
else:
# this is just for the layout if no nvidia_gpu is found.
gpu_stats = [""]
# Create the "Built-in RAM" label, slider, and value label
self.ram_label = QLabel("Built-in RAM:")
ram_info = psutil.virtual_memory()
total_ram = ram_info.total // (1024 ** 2) # Convert to MiB
used_ram = ram_info.used // (1024 ** 2) # Convert to MiB
free_ram = ram_info.available // (1024 ** 2) # Convert to MiB
self.ram_label.setToolTip(f"Total RAM: {total_ram} MiB\nUsed RAM: {used_ram} MiB\nFree RAM: {free_ram} MiB")
self.ram_label.hide()
layout.addWidget(self.ram_label, 11, 0)
self.ram_slider = QSlider(Qt.Horizontal)
self.ram_slider.setMinimum(0)
self.ram_slider.setMaximum(100)
self.ram_slider.setTickInterval(1)
self.ram_slider.setSingleStep(1)
self.ram_slider.hide()
layout.addWidget(self.ram_slider, 11, 1)
self.ram_value_label = QLabel("0 GiB")
self.ram_value_label.hide()
layout.addWidget(self.ram_value_label, 11, 2)
# Connect the valueChanged signal of the RAM slider to update the value label
self.ram_slider.valueChanged.connect(self.on_ram_slider_changed)
# Pre-layer Slider
self.pre_layer_labels = []
self.pre_layer_slider = []
self.pre_layer_slider_value = []
self.pre_layer_amount_max = 100
# 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.
if nvidia_gpu:
for i, gpu in enumerate(gpu_stats):
pre_layer_labels = QLabel(f"{gpu.name} Pre_Layer:")
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.")
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.setMaximum(100)
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)
self.pre_layer_slider.append(pre_layer_sliders)
pre_layer_sliders_value = QLabel("0")
layout.addWidget(pre_layer_sliders_value, 11 + (len(gpu_stats) * 2) + i, 2)
self.pre_layer_slider_value.append(pre_layer_sliders_value)
# Add horizontal line to seperate the Checkboxes
line = QFrame()
line.setFrameShape(QFrame.HLine)
line.setFrameShadow(QFrame.Sunken)
layout.addWidget(line, 13 + (len(gpu_stats) * 2), 0, 1, 3)
# Load in 8 Bit Mode
self.use_8bit_checkbox = QCheckBox("Load in 8bit")
self.use_8bit_checkbox.setToolTip("VRAM Reducing!\nReduces memory usage and computational complexity at the cost of lower precision compared to higher precision representations.")
layout.addWidget(self.use_8bit_checkbox, 14 + (len(gpu_stats) * 2), 0)
# Deactivate Streaming Output
self.use_nostream_checkbox = QCheckBox("No Stream")
self.use_nostream_checkbox.setToolTip("Don't stream the text output in real time. Increases Token/s by ~ 50%")
layout.addWidget(self.use_nostream_checkbox, 14 + (len(gpu_stats) * 2), 1)
# Load in full 16bit precision
self.use_16bit_checkbox = QCheckBox("Load in 16bit")
self.use_16bit_checkbox.setToolTip("Load the model with bfloat16 precision. Requires NVIDIA Ampere GPU.")
layout.addWidget(self.use_16bit_checkbox, 15 + (len(gpu_stats) * 2), 0)
# Use xformers
self.use_xformers_checkbox = QCheckBox("xformers")
self.use_xformers_checkbox.setToolTip("Use xformer's memory efficient attention. This should increase your tokens/s.")
layout.addWidget(self.use_xformers_checkbox, 15 + (len(gpu_stats) * 2), 1)
# Make use of Remote Code Execution (MPT-7B)
self.use_trc_checkbox = QCheckBox("trust-remote-code")
self.use_trc_checkbox.setToolTip("Set trust_remote_code=True while loading a model. Necessary for ChatGLM and MPT-7B.")
layout.addWidget(self.use_trc_checkbox, 16 + (len(gpu_stats) * 2), 0)
# Load with Monkey-Patch enabled
self.use_monkey_checkbox = QCheckBox("Monkey Patch")
self.use_monkey_checkbox.setToolTip("Apply the monkey patch for using LoRAs with quantized models.")
layout.addWidget(self.use_monkey_checkbox, 16 + (len(gpu_stats) * 2), 1)
# Use Triton Quant-ATTN
self.use_quant_checkbox = QCheckBox("Quant_attn")
self.use_quant_checkbox.setToolTip("(triton) Enable quant attention.")
layout.addWidget(self.use_quant_checkbox, 17 + (len(gpu_stats) * 2), 0)
# Use Triton Warmup & Autotune
self.use_autotune_checkbox = QCheckBox("Warmup-Autotune")
self.use_autotune_checkbox.setToolTip("(triton) Enable warmup autotune.")
layout.addWidget(self.use_autotune_checkbox, 17 + (len(gpu_stats) * 2), 1)
# Disable Cache for better VRAM
self.use_nocache_checkbox = QCheckBox("No Cache")
self.use_nocache_checkbox.setToolTip("VRAM Reducing!\nSet use_cache to False while generating text. This reduces the VRAM usage a bit with a performance cost.")
layout.addWidget(self.use_nocache_checkbox, 18 + (len(gpu_stats) * 2), 1)
# Use DISK to load part of the model
self.use_disk_checkbox = QCheckBox("Use DISK")
self.use_disk_checkbox.setToolTip("If the model is too large for your GPU(s) and CPU combined, send the remaining layers to the disk.")
layout.addWidget(self.use_disk_checkbox, 18 + (len(gpu_stats) * 2), 0)
self.use_disk_checkbox.stateChanged.connect(self.on_use_disk_checkbox_changed)
# Use DISK to load part of the model
self.change_disk_cache_checkbox = QCheckBox("Change Disk Cache")
self.change_disk_cache_checkbox.setVisible(False)
self.change_disk_cache_checkbox.setToolTip("OPTIONAL: Change the disk cache directory.")
layout.addWidget(self.change_disk_cache_checkbox, 19 + (len(gpu_stats) * 2), 0)
self.change_disk_cache_checkbox.stateChanged.connect(self.on_change_disk_cache_checkbox_changed)
# Current Cache Folder
self.disk_cache_path = ""
self.current_disk_cache_label = QLabel("Current Cache Folder:")
self.current_disk_cache_label.setVisible(False)
self.current_disk_cache_label.setToolTip("The current disk cache folder.")
layout.addWidget(self.current_disk_cache_label, 19 + (len(gpu_stats) * 2), 1)
# Choose Folder Field
self.choose_disk_folder_label = QLabel("Choose Folder:")
self.choose_disk_folder_label.setVisible(False)
self.choose_disk_folder_label.setToolTip("Choose a folder to use for the disk cache.")
layout.addWidget(self.choose_disk_folder_label, 20 + (len(gpu_stats) * 2), 0)
# Choose Folder button
self.choose_disk_folder_button = QPushButton("Choose Folder")
self.choose_disk_folder_button.setVisible(False)
self.disk_cache_textfield = QLineEdit()
self.choose_disk_folder_button.setToolTip("Choose a folder to use for the disk cache.")
self.choose_disk_folder_button.clicked.connect(self.on_choose_disk_folder_button_clicked)
layout.addWidget(self.choose_disk_folder_button, 20 + (len(gpu_stats) * 2), 1)
# Use sdp_attention
self.use_sdp_attention_checkbox = QCheckBox("Use sdp-attention")
self.use_sdp_attention_checkbox.setToolTip("Use torch 2.0's sdp attention.")
layout.addWidget(self.use_sdp_attention_checkbox, 21 + (len(gpu_stats) * 2), 0)
# Use Multimodal Checkbox
self.use_multimodal_checkbox = QCheckBox("Multimodal")
self.use_multimodal_checkbox.setToolTip("Use multimodal models.")
layout.addWidget(self.use_multimodal_checkbox, 21 + (len(gpu_stats) * 2), 1)
# Add autogptq checkbox
self.use_autogptq_checkbox = QCheckBox("AutoGPTQ")
self.use_autogptq_checkbox.setToolTip("Use AutoGPTQ for loading quantized models instead of the internal GPTQ loader.")
layout.addWidget(self.use_autogptq_checkbox, 22 + (len(gpu_stats) * 2), 0)
# Add Triton checkbox
self.use_triton_checkbox = QCheckBox("Triton")
self.use_triton_checkbox.setToolTip("Use Triton for inference.")
layout.addWidget(self.use_triton_checkbox, 22 + (len(gpu_stats) * 2), 1)
# Add horizontal line to seperate the Checkboxes
line = QFrame()
line.setFrameShape(QFrame.HLine)
line.setFrameShadow(QFrame.Sunken)
layout.addWidget(line, 23 + (len(gpu_stats) * 2), 0, 1, 3)
# New GUI Options based on Toolbox Checkboxes.
######################################################
# ____ ____ _ #
# | _ \ ___ ___ _ __/ ___| _ __ ___ ___ __| | #
# | | | |/ _ \/ _ \ '_ \___ \| '_ \ / _ \/ _ \/ _` | #
# | |_| | __/ __/ |_) |__) | |_) | __/ __/ (_| | #
# |____/ \___|\___| .__/____/| .__/ \___|\___|\__,_| #
# |_| |_| #
######################################################
# Deepspeed Header
self.deepspeed_label_header = QLabel("Deepspeed Options:")
self.deepspeed_label_header.setToolTip("Deepspeed Options")
layout.addWidget(self.deepspeed_label_header, 30 + (len(gpu_stats) * 2), 0)
self.deepspeed_label_header.setVisible(False)
# Deepspeed Checkbox
self.deepspeed_checkbox = QCheckBox("Use Deepspeed")
self.deepspeed_checkbox.setToolTip("Enable the use of DeepSpeed ZeRO-3 for inference via the Transformers integration.")
layout.addWidget(self.deepspeed_checkbox, 31 + (len(gpu_stats) * 2), 0)
self.deepspeed_checkbox.setVisible(False)
# Deepspeed Box
deepspeed_box = QHBoxLayout()
# Deepspeed GPU num Label
self.deepspeed_gpu_num_label = QLabel("Deepspeed GPU num:")
self.deepspeed_gpu_num_label.setVisible(False)
self.deepspeed_gpu_num_label.setToolTip("The number of GPUs to use for DeepSpeed ZeRO-3.")
deepspeed_box.addWidget(self.deepspeed_gpu_num_label)
# Deepspeed GPU num Spinbox
self.deepspeed_gpu_num_spinbox = QSpinBox()
self.deepspeed_gpu_num_spinbox.setVisible(False)
self.deepspeed_gpu_num_spinbox.setToolTip("The number of GPUs to use for DeepSpeed ZeRO-3.")
self.deepspeed_gpu_num_spinbox.setMinimum(1)
self.deepspeed_gpu_num_spinbox.setMaximum(16)
deepspeed_box.addWidget(self.deepspeed_gpu_num_spinbox)
layout.addLayout(deepspeed_box, 31 + (len(gpu_stats) * 2), 1, 1, 2)
# Deepspeed use NVMe Offload Directory Checkbox
self.deepspeed_nvme_checkbox = QCheckBox("Use Offload Directory")
self.deepspeed_nvme_checkbox.setVisible(False)
self.deepspeed_nvme_checkbox.setToolTip("Use an NVMe offload directory for ZeRO-3.")
layout.addWidget(self.deepspeed_nvme_checkbox, 32 + (len(gpu_stats) * 2), 0)
self.deepspeed_nvme_checkbox.stateChanged.connect(self.on_deepspeed_nvme_checkbox_changed)
# NVMe offload Directory
self.deepspeed_nvme_label = QLabel("NVMe Offload Directory:")
self.deepspeed_nvme_label.setVisible(False)
self.deepspeed_nvme_label.setToolTip("Directory to use for ZeRO-3 NVME offloading.")
layout.addWidget(self.deepspeed_nvme_label, 33 + (len(gpu_stats) * 2), 0)
# NVMe Current Offload Directory
self.selected_offload_directory = "none"
self.deepspeed_nvme_current_label = QLabel(f"Current Directory: {self.selected_offload_directory}")
self.deepspeed_nvme_current_label.setVisible(False)
self.deepspeed_nvme_current_label.setToolTip("The current NVMe offload directory.")
layout.addWidget(self.deepspeed_nvme_current_label, 33 + (len(gpu_stats) * 2), 1)
# NVMe Offload Directory folder choose
self.deepspeed_nvme_button = QPushButton("Choose Folder")
self.deepspeed_nvme_button.setVisible(False)
self.deepspeed_nvme_button.setToolTip("Choose a folder to use for the NVMe offload.")
self.deepspeed_nvme_button.clicked.connect(self.on_deepspeed_nvme_button_clicked)
layout.addWidget(self.deepspeed_nvme_button, 34 + (len(gpu_stats) * 2), 1)
# Local Rank
self.deepspeed_local_rank_label = QLabel("Local Rank:")
self.deepspeed_local_rank_label.setVisible(False)
self.deepspeed_local_rank_label.setToolTip("Optional argument for distributed setups.")
layout.addWidget(self.deepspeed_local_rank_label, 35 + (len(gpu_stats) * 2), 0)
# Local Rank SpinBox
self.deepspeed_local_rank_spinbox = QSpinBox()
self.deepspeed_local_rank_spinbox.setVisible(False)
self.deepspeed_local_rank_spinbox.setToolTip("Optional argument for distributed setups.")
self.deepspeed_local_rank_spinbox.setMinimum(0)
layout.addWidget(self.deepspeed_local_rank_spinbox, 35 + (len(gpu_stats) * 2), 1)
# Add horizontal line to seperate the Checkboxes
self.deepspeed_line = QFrame()
self.deepspeed_line.setFrameShape(QFrame.HLine)
self.deepspeed_line.setFrameShadow(QFrame.Sunken)
self.deepspeed_line.setVisible(False)
layout.addWidget(self.deepspeed_line, 36 + (len(gpu_stats) * 2), 0, 1, 3)
#################################################
# _ _ #
# | | | __ _ _ __ ___ __ _ ___ _ __ _ __ #
# | | |/ _` | '_ ` _ \ / _` | / __| '_ \| '_ \ #
# | | | (_| | | | | | | (_| || (__| |_) | |_) | #
# |_|_|\__,_|_| |_| |_|\__,_(_)___| .__/| .__/ #
# |_| |_| #
#################################################
# llama.cpp Header
self.llama_label_header = QLabel("llama.cpp Options:")
self.llama_label_header.setToolTip("llama.cpp Options")
layout.addWidget(self.llama_label_header, 40 + (len(gpu_stats) * 2), 0)
self.llama_label_header.setVisible(False)
# llama.cpp threads box
llama_threads_box = QHBoxLayout()
# llama.cpp threads
self.llama_threads_label = QLabel("Threads:")
self.llama_threads_label.setVisible(False)
self.llama_threads_label.setToolTip("Number of threads to use for llama.cpp.")
llama_threads_box.addWidget(self.llama_threads_label)
# llama.cpp threads number
self.llama_threads_spinbox = QSpinBox()
self.llama_threads_spinbox.setToolTip("Number of threads to use for llama.cpp.")
self.llama_threads_spinbox.setRange(0, max_threads)
self.llama_threads_spinbox.setValue(0) # Set an initial value
self.llama_threads_spinbox.setVisible(False)
llama_threads_box.addWidget(self.llama_threads_spinbox)
layout.addLayout(llama_threads_box, 42 + (len(gpu_stats) * 2), 0)
# llama.cpp batch size box
llama_batch_size_box = QHBoxLayout()
# llama.cpp batch size
self.llama_batch_size_label = QLabel("Batch Size:")
self.llama_batch_size_label.setVisible(False)
self.llama_batch_size_label.setToolTip("Maximum number of prompt tokens to batch together when calling llama_eval.")
llama_batch_size_box.addWidget(self.llama_batch_size_label)
# llama.cpp batch size number
self.llama_batch_size_spinbox = QSpinBox()
self.llama_batch_size_spinbox.setToolTip("Maximum number of prompt tokens to batch together when calling llama_eval.")
self.llama_batch_size_spinbox.setRange(4, 8192)
self.llama_batch_size_spinbox.setValue(512)
self.llama_batch_size_spinbox.setSingleStep(4)
self.llama_batch_size_spinbox.setVisible(False)
llama_batch_size_box.addWidget(self.llama_batch_size_spinbox)
layout.addLayout(llama_batch_size_box, 42 + (len(gpu_stats) * 2), 1, 1, 2)
# llama.cpp mmap checkbox
self.llama_mmap_checkbox = QCheckBox("Use no mmap")
self.llama_mmap_checkbox.setToolTip("Prevent mmap from being used.")
layout.addWidget(self.llama_mmap_checkbox, 44 + (len(gpu_stats) * 2), 0)
self.llama_mmap_checkbox.setVisible(False)
# llama mlock checkbox
self.llama_mlock_checkbox = QCheckBox("Use mlock")
self.llama_mlock_checkbox.setToolTip("Force the system to keep the model in RAM.")
layout.addWidget(self.llama_mlock_checkbox, 44 + (len(gpu_stats) * 2), 1)
self.llama_mlock_checkbox.setVisible(False)
# llama.cpp cache capacity box
llama_cache_capacity_box = QHBoxLayout()
# llama Cache Capacity Spinbox
self.llama_cache_capacity_label = QLabel("Cache Capacity:")
self.llama_cache_capacity_label.setVisible(False)
self.llama_cache_capacity_label.setToolTip("Maximum number of prompt tokens to cache in RAM.")
llama_cache_capacity_box.addWidget(self.llama_cache_capacity_label)
# llama Cache Capacity Spinbox
self.llama_cache_capacity_spinbox = QSpinBox()
self.llama_cache_capacity_spinbox.setToolTip("Maximum number of prompt tokens to cache in RAM.")
self.llama_cache_capacity_spinbox.setRange(0, 8192)
self.llama_cache_capacity_spinbox.setValue(1024)
self.llama_cache_capacity_spinbox.setSingleStep(4)
self.llama_cache_capacity_spinbox.setVisible(False)
llama_cache_capacity_box.addWidget(self.llama_cache_capacity_spinbox)
# llama.cpp Cache Capacity Units
self.llama_cache_capacity_units = QComboBox()
self.llama_cache_capacity_units.setToolTip("Choose the Units to use")
self.llama_cache_capacity_units.addItems(["MiB", "GiB"])
self.llama_cache_capacity_units.setVisible(False)
llama_cache_capacity_box.addWidget(self.llama_cache_capacity_units)
layout.addLayout(llama_cache_capacity_box, 45 + (len(gpu_stats) * 2), 0)
# GPU Layer Box
self.llama_gpu_layer_box = QHBoxLayout()
# llama GPU Layer Label
self.llama_gpu_layer_label = QLabel("GPU Layer:")
self.llama_gpu_layer_label.setVisible(False)
self.llama_gpu_layer_label.setToolTip("Number of layers to offload to the GPU.")
self.llama_gpu_layer_box.addWidget(self.llama_gpu_layer_label)
# llama GPU Layer Number
self.llama_gpu_layer_spinbox = QSpinBox()
self.llama_gpu_layer_spinbox.setToolTip("Number of layers to offload to the GPU.\nTo Offload all to GPU set it to 200.000 (MAX)")
self.llama_gpu_layer_spinbox.setRange(0, 200000)
self.llama_gpu_layer_spinbox.setValue(0)
self.llama_gpu_layer_spinbox.setSingleStep(1)
self.llama_gpu_layer_spinbox.setVisible(False)
self.llama_gpu_layer_box.addWidget(self.llama_gpu_layer_spinbox)
layout.addLayout(self.llama_gpu_layer_box, 45 + (len(gpu_stats) * 2), 1, 1, 2)
# llama.cpp n_ctx inner layout
llama_n_ctx_inner_layout = QHBoxLayout()
# llama.cpp n_ctx label
self.llama_n_ctx_label = QLabel("n_ctx:")
self.llama_n_ctx_label.setVisible(False)
self.llama_n_ctx_label.setToolTip("Size of the prompt context.")
llama_n_ctx_inner_layout.addWidget(self.llama_n_ctx_label)
# llama.cpp n_ctx size dropdown
self.llama_n_ctx_dropdown = QComboBox()
self.llama_n_ctx_dropdown.setToolTip("Size of the prompt context.")
self.llama_n_ctx_dropdown.addItems(["128", "256", "512", "1024", "2048", "4096", "8192"])
self.llama_n_ctx_dropdown.setCurrentIndex(4)
self.llama_n_ctx_dropdown.setVisible(False)
llama_n_ctx_inner_layout.addWidget(self.llama_n_ctx_dropdown)
layout.addLayout(llama_n_ctx_inner_layout, 46 + (len(gpu_stats) * 2), 0)
# llama.cpp seed layout
llama_seed_inner_layout = QHBoxLayout()
# llama.cpp seed label
self.llama_seed_label = QLabel("Seed:")
self.llama_seed_label.setVisible(False)
self.llama_seed_label.setToolTip("Seed for llama-cpp models. Default 0 (random).")
llama_seed_inner_layout.addWidget(self.llama_seed_label)
# llama.cpp seed spinbox
self.llama_seed_spinbox = QSpinBox()
self.llama_seed_spinbox.setToolTip("Seed for llama-cpp models. Default 0 (random).")
self.llama_seed_spinbox.setRange(0, 2147483647)
self.llama_seed_spinbox.setValue(0)
self.llama_seed_spinbox.setSingleStep(1)
self.llama_seed_spinbox.setVisible(False)
llama_seed_inner_layout.addWidget(self.llama_seed_spinbox)
layout.addLayout(llama_seed_inner_layout, 46 + (len(gpu_stats) * 2), 1, 1, 2)
# Seperator for the Toolbox Options
self.llama_line = QFrame()
self.llama_line.setFrameShape(QFrame.HLine)
self.llama_line.setFrameShadow(QFrame.Sunken)
self.llama_line.setVisible(False)
layout.addWidget(self.llama_line, 49 + (len(gpu_stats) * 2), 0, 1, 3)
########################################
# _____ _ ____ #
# | ___| | _____ __/ ___| ___ _ __ #
# | |_ | |/ _ \ \/ / | _ / _ \ '_ \ #
# | _| | | __/> <| |_| | __/ | | | #
# |_| |_|\___/_/\_\\____|\___|_| |_| #
# #
########################################
# FlexGen Header Label
self.flexgen_header_label = QLabel("FlexGen Options")
self.flexgen_header_label.setVisible(False)
self.flexgen_header_label.setToolTip("Options for FlexGen.")
layout.addWidget(self.flexgen_header_label, 50 + (len(gpu_stats) * 2), 0)
# FlexGen Checkbox
self.flexgen_checkbox = QCheckBox("Use FlexGen")
self.flexgen_checkbox.setToolTip("Enable the use of FlexGen offloading.")
self.flexgen_checkbox.setVisible(False)
layout.addWidget(self.flexgen_checkbox, 50 + (len(gpu_stats) * 2), 0)
#self.flexgen_checkbox.stateChanged.connect(self.on_flexgen_checkbox_changed)
# FlexGen Percentage
inner_layout = QHBoxLayout()
self.flexgen_percentage_label = QLabel("FlexGen Percentage:")
#self.flexgen_percentage_label.setVisible(False)
self.flexgen_percentage_label.setToolTip("FlexGen: allocation percentages. Must be 6 numbers separated by spaces (default: 0, 100, 100, 0, 100, 0).\n\nthe percentage of weight on GPU\nthe percentage of weight on CPU\nthe percentage of attention cache on GPU\nthe percentage of attention cache on CPU\nthe percentage of activations on GPU\nthe percentage of activations on CPU")
self.flexgen_percentage_label.setVisible(False)
inner_layout.addWidget(self.flexgen_percentage_label)
# FlexGen Percentage Spinbox 1
self.flexgen_percentage_spinbox1 = QSpinBox()
self.flexgen_percentage_spinbox1.setToolTip("Allocation percentages. Default: 0\nthe percentage of weight on GPU")
self.flexgen_percentage_spinbox1.setRange(0, 100)
self.flexgen_percentage_spinbox1.setValue(0)
self.flexgen_percentage_spinbox1.setVisible(False)
inner_layout.addWidget(self.flexgen_percentage_spinbox1)
# FlexGen Percentage Spinbox 2
self.flexgen_percentage_spinbox2 = QSpinBox()
self.flexgen_percentage_spinbox2.setToolTip("Allocation percentages. Default: 100.\nthe percentage of weight on CPU")
self.flexgen_percentage_spinbox2.setRange(0, 100)
self.flexgen_percentage_spinbox2.setValue(100)
self.flexgen_percentage_spinbox2.setVisible(False)
inner_layout.addWidget(self.flexgen_percentage_spinbox2)
# FlexGen Percentage Spinbox 3
self.flexgen_percentage_spinbox3 = QSpinBox()
self.flexgen_percentage_spinbox3.setToolTip("Allocation percentages. Default: 100.\nthe percentage of attention cache on GPU")
self.flexgen_percentage_spinbox3.setRange(0, 100)
self.flexgen_percentage_spinbox3.setValue(100)
self.flexgen_percentage_spinbox3.setVisible(False)
inner_layout.addWidget(self.flexgen_percentage_spinbox3)
# FlexGen Percentage Spinbox 4
self.flexgen_percentage_spinbox4 = QSpinBox()
self.flexgen_percentage_spinbox4.setToolTip("Allocation percentages. Default: 0.\nthe percentage of attention cache on CPU")
self.flexgen_percentage_spinbox4.setRange(0, 100)
self.flexgen_percentage_spinbox4.setValue(0)
self.flexgen_percentage_spinbox4.setVisible(False)
inner_layout.addWidget(self.flexgen_percentage_spinbox4)
# FlexGen Percentage Spinbox 5
self.flexgen_percentage_spinbox5 = QSpinBox()
self.flexgen_percentage_spinbox5.setToolTip("Allocation percentages. Default: 100.\nthe percentage of activations on GPU")
self.flexgen_percentage_spinbox5.setRange(0, 100)
self.flexgen_percentage_spinbox5.setValue(100)
self.flexgen_percentage_spinbox5.setVisible(False)
inner_layout.addWidget(self.flexgen_percentage_spinbox5)
# FlexGen Percentage Spinbox 6
self.flexgen_percentage_spinbox6 = QSpinBox()
self.flexgen_percentage_spinbox6.setToolTip("Allocation percentages. Default: 0.\nthe percentage of activations on CPU")
self.flexgen_percentage_spinbox6.setRange(0, 100)
self.flexgen_percentage_spinbox6.setValue(0)
self.flexgen_percentage_spinbox6.setVisible(False)
inner_layout.addWidget(self.flexgen_percentage_spinbox6)
layout.addLayout(inner_layout, 51 + (len(gpu_stats) * 2), 0, 1, 3)
# FlexGen compression Checkbox
self.flexgen_compression_checkbox = QCheckBox("Use Compression")
self.flexgen_compression_checkbox.setToolTip("Enable the use of compression for FlexGen.")
self.flexgen_compression_checkbox.setVisible(False)
#self.flexgen_compression_checkbox.stateChanged.connect(self.on_flexgen_compression_checkbox_changed)
layout.addWidget(self.flexgen_compression_checkbox, 52 + (len(gpu_stats) * 2), 0)
# FlexGen pin weight QLabel
self.flexgen_pin_weight_label = QLabel("FlexGen pin weight:")
self.flexgen_pin_weight_label.setVisible(False)
self.flexgen_pin_weight_label.setToolTip("Pin weight for FlexGen. Default: 0.")
layout.addWidget(self.flexgen_pin_weight_label, 53 + (len(gpu_stats) * 2), 0)
# FlexGen pin weight dropdown
self.flexgen_pin_weight_dropdown = QComboBox()
self.flexgen_pin_weight_dropdown.setToolTip("Pin weight for FlexGen. Default: 0.")
self.flexgen_pin_weight_dropdown.setVisible(False)
self.flexgen_pin_weight_dropdown.addItems(["none", "True", "False"])
self.flexgen_pin_weight_dropdown.setCurrentIndex(0)
layout.addWidget(self.flexgen_pin_weight_dropdown, 53 + (len(gpu_stats) * 2), 1)
# Seperator for the Toolbox Options
self.flexline = QFrame()
self.flexline.setFrameShape(QFrame.HLine)
self.flexline.setFrameShadow(QFrame.Sunken)
self.flexline.setVisible(False)
layout.addWidget(self.flexline, 54 + (len(gpu_stats) * 2), 0, 1, 3)
###################################
# ______ ___ ____ __ #
# | _ \ \ / / |/ /\ \ / / #
# | |_) \ \ /\ / /| ' / \ \ / / #
# | _ < \ V V / | . \ \ V / #
# |_| \_\ \_/\_/ |_|\_\ \_/ #
# #
###################################
# RWKV Header
self.rwkv_header = QLabel("RWKV:")
self.rwkv_header.setVisible(False)
self.rwkv_header.setToolTip("RWKV: allocation percentages. Must be 6 numbers separated by spaces (default: 0, 100, 100, 0, 100, 0).")
layout.addWidget(self.rwkv_header, 60 + (len(gpu_stats) * 2), 0)
# RWKV Checkbox
self.rwkv_checkbox = QCheckBox("Enable RWKV")
self.rwkv_checkbox.setToolTip("Enable RWKV.")
self.rwkv_checkbox.setVisible(False)
layout.addWidget(self.rwkv_checkbox, 61 + (len(gpu_stats) * 2), 0)
#self.rwkv_checkbox.stateChanged.connect(self.on_rwkv_checkbox_changed)
# RWKV Strategy Checkbox
self.rwkv_strategy_checkbox = QCheckBox("Enable RWKV Strategy")
self.rwkv_strategy_checkbox.setToolTip("Enable RWKV Strategy.")
self.rwkv_strategy_checkbox.setVisible(False)
layout.addWidget(self.rwkv_strategy_checkbox, 62 + (len(gpu_stats) * 2), 0)
#self.rwkv_strategy_checkbox.stateChanged.connect(self.on_rwkv_strategy_checkbox_changed)
# RWKV Strategy dropdown
rwkv_horizontalbox = QHBoxLayout()
# Dropdown for the Strategy Modes
self.rwkv_strategy_dropdown = QComboBox()
self.rwkv_strategy_dropdown.setToolTip("If you want to use a specific RWKV Strategy, you can choose here to enter which mode and the strategy strength\"cpu fp32\" # CPU mode\n\"cuda fp16\" # GPU mode with float16 precision\n\"cuda fp16 *30 -> cpu fp32\" # GPU+CPU offloading. The higher the number after *, the higher the GPU allocation.\n\"cuda fp16i8\" # GPU mode with 8-bit precision")
self.rwkv_strategy_dropdown.setVisible(False)
self.rwkv_strategy_dropdown.addItems(["none", "cpu fp32", "cuda fp16", "cuda fp16i8"])
self.rwkv_strategy_dropdown.setCurrentIndex(0)
rwkv_horizontalbox.addWidget(self.rwkv_strategy_dropdown)
# RWKV Allocation Spinbox
self.rwkv_allocation_spinbox = QSpinBox()
self.rwkv_allocation_spinbox.setToolTip("If you want to use a specific RWKV Strategy, you can choose here to enter which mode and the strategy strength\"cpu fp32\" # CPU mode\n\"cuda fp16\" # GPU mode with float16 precision\n\"cuda fp16 *30 -> cpu fp32\" # GPU+CPU offloading. The higher the number after *, the higher the GPU allocation.\n\"cuda fp16i8\" # GPU mode with 8-bit precision")
self.rwkv_allocation_spinbox.setVisible(False)
self.rwkv_allocation_spinbox.setRange(0, 100)
self.rwkv_allocation_spinbox.setValue(0)
rwkv_horizontalbox.addWidget(self.rwkv_allocation_spinbox)
layout.addLayout(rwkv_horizontalbox, 62 + (len(gpu_stats) * 2), 1, 1 ,2)
# RWKV Cuda Checkbox
self.rwkv_cuda_checkbox = QCheckBox("Enable RWKV Cuda")
self.rwkv_cuda_checkbox.setToolTip("Enable RWKV Cuda.")
self.rwkv_cuda_checkbox.setVisible(False)
layout.addWidget(self.rwkv_cuda_checkbox, 64 + (len(gpu_stats) * 2), 0)
# Seperator for the RWKV
self.rwkv_line = QFrame()
self.rwkv_line.setFrameShape(QFrame.HLine)
self.rwkv_line.setFrameShadow(QFrame.Sunken)
self.rwkv_line.setVisible(False)
layout.addWidget(self.rwkv_line, 65 + (len(gpu_stats) * 2), 0, 1, 3)
######################
# _ ____ ___ #
# / \ | _ \_ _| #
# / _ \ | |_) | | #
# / ___ \| __/| | #
# /_/ \_\_| |___| #
# #
######################
# API Header Label
self.api_header = QLabel("API:")
self.api_header.setVisible(False)
self.api_header.setToolTip("API: Choose the API settings to use.")
layout.addWidget(self.api_header, 70 + (len(gpu_stats) * 2), 0)
# API Checkbox
self.api_checkbox = QCheckBox("Enable API")
self.api_checkbox.setToolTip("Enable the API extension.")
self.api_checkbox.setVisible(False)
layout.addWidget(self.api_checkbox, 71 + (len(gpu_stats) * 2), 0)
# API blocking Port Checkbox
self.api_blocking_port_checkbox = QCheckBox("Change API Blocking Port")
self.api_blocking_port_checkbox.setToolTip("The listening port for the blocking API.\nDefault: 5000")
self.api_blocking_port_checkbox.setVisible(False)
layout.addWidget(self.api_blocking_port_checkbox, 72 + (len(gpu_stats) * 2), 0)
self.api_blocking_port_checkbox.stateChanged.connect(self.on_api_blocking_port_checkbox_changed)
# API Blocking Port Spinbox
self.api_blocking_port_SpinBox = QSpinBox()
self.api_blocking_port_SpinBox.setToolTip("The listening port for the blocking API.\nDefault: 5000")
self.api_blocking_port_SpinBox.setVisible(False)
self.api_blocking_port_SpinBox.setEnabled(False)
self.api_blocking_port_SpinBox.setRange(0, 65535)
self.api_blocking_port_SpinBox.setValue(5000)
layout.addWidget(self.api_blocking_port_SpinBox, 72 + (len(gpu_stats) * 2), 1)
# API Streaming Port Checkbox
self.api_streaming_port_checkbox = QCheckBox("Change API Streaming Port")
self.api_streaming_port_checkbox.setToolTip("The listening port for the streaming API.\nDefault: 5005")
self.api_streaming_port_checkbox.setVisible(False)
layout.addWidget(self.api_streaming_port_checkbox, 73 + (len(gpu_stats) * 2), 0)
self.api_streaming_port_checkbox.stateChanged.connect(self.on_api_streaming_port_checkbox_changed)
# API Streaming Port Textfield
self.api_streaming_port_SpinBox = QSpinBox()
self.api_streaming_port_SpinBox.setToolTip("The listening port for the streaming API.\nDefault: 5005")
self.api_streaming_port_SpinBox.setVisible(False)
self.api_streaming_port_SpinBox.setEnabled(False)
self.api_streaming_port_SpinBox.setRange(0, 65535)
self.api_streaming_port_SpinBox.setValue(5005)
layout.addWidget(self.api_streaming_port_SpinBox, 73 + (len(gpu_stats) * 2), 1)
# Enable Public API
self.api_public_checkbox = QCheckBox("Enable Public API")
self.api_public_checkbox.setToolTip("Create a public URL for the API using Cloudfare.")
self.api_public_checkbox.setVisible(False)
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, 84 + (len(gpu_stats) * 2), 0, 1, 3)
# Authentication Box
authentication_box = QHBoxLayout()
# Authentication Checkbox
self.authentication_checkbox = QCheckBox("Authentication")
self.authentication_checkbox.setToolTip("Enable gradio authentication")
self.authentication_checkbox.stateChanged.connect(self.on_authentication_checkbox_changed)
authentication_box.addWidget(self.authentication_checkbox)
# Choose File Field
self.choose_file_label = QLabel("Choose File:")
self.choose_file_label.setVisible(False)
self.choose_file_label.setToolTip("Choose a file to use for the authentication credentials. Credentials should be saved like:\nUSERNAME1:PASSWORD1\nUSERNAME2:PASSWORD2")
authentication_box.addWidget(self.choose_file_label)
self.choose_file_button = QPushButton("Browse")
self.choose_file_button.setVisible(False)
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, 85 + (len(gpu_stats) * 2), 0, 1, 3)
# Extensions Selection Menu
self.use_extensions_checkbox = QCheckBox("Use Extensions")
self.use_extensions_checkbox.setToolTip("Choose the extensions to be loaded.")
layout.addWidget(self.use_extensions_checkbox, 90 + (len(gpu_stats) * 2), 0)
self.use_extensions_checkbox.stateChanged.connect(self.on_use_extensions_checkbox_changed)
self.extensions_list = QListWidget()
self.extensions_list.setToolTip("Choose the extensions to be loaded.")
layout.addWidget(self.extensions_list, 90 + (len(gpu_stats) * 2), 1, 1, 2)
self.extensions_list.setFixedHeight(150)
self.extensions_list.setVisible(False)
extensions = [name for name in os.listdir(extensions_folder) if os.path.isdir(os.path.join(extensions_folder, name)) and "api" not in name.lower()]
extensions.sort()
for extension in extensions:
item = QListWidgetItem(extension)
item.setFlags(item.flags() | Qt.ItemIsUserCheckable)
item.setCheckState(Qt.Unchecked)
self.extensions_list.addItem(item)
# Lora selection menu
self.use_lora_checkbox = QCheckBox("Use Loras")
self.use_lora_checkbox.setToolTip("Choose the loras to be loaded.")
layout.addWidget(self.use_lora_checkbox, 100 + (len(gpu_stats) * 2), 0)
self.use_lora_checkbox.stateChanged.connect(self.on_use_lora_checkbox_changed)
self.lora_list = QListWidget()
self.lora_list.setToolTip("Choose the loras to be loaded.")
layout.addWidget(self.lora_list, 100 + (len(gpu_stats) * 2), 1, 1, 2)
self.lora_list.setVisible(False)
loras = [name for name in os.listdir(loras_folder) if os.path.isdir(os.path.join(loras_folder, name))]
for lora in loras:
item = QListWidgetItem(lora)
item.setFlags(item.flags() | Qt.ItemIsUserCheckable)
item.setCheckState(Qt.Unchecked)
self.lora_list.addItem(item)
# Use Whole Local Network
self.use_network_checkbox = QCheckBox("Local Network Mode")
self.use_network_checkbox.setToolTip("By default, the WebUI will only be reachable by the PC running it.\nIf you want to use it also on other devices, check this")
layout.addWidget(self.use_network_checkbox, 110 + (len(gpu_stats) * 2), 0)
# Use Automatically opens the Browser when finished loading the webui
self.use_autolaunch_checkbox = QCheckBox("Auto open Browser")
self.use_autolaunch_checkbox.setToolTip("Automatically Opens your browser when loading is finished")
layout.addWidget(self.use_autolaunch_checkbox, 120 + (len(gpu_stats) * 2), 0)
# Listen Port Checkbox and Text Field
self.listen_port_checkbox = QCheckBox("Listen Port")
self.listen_port_checkbox.setToolTip("Choose the Port to use for the WebUI.\nDefault is 7680. If you want to use Stable Diffusion at the same time,\nor got other services running on this Port, you can change it in the textfield.")
self.listen_port_checkbox.stateChanged.connect(self.on_listen_port_checkbox_changed)
layout.addWidget(self.listen_port_checkbox, 130 + (len(gpu_stats) * 2), 1)
# Auto Close the GUI when pressing start.
self.use_autoclose_checkbox = QCheckBox("Close GUI on Start")
self.use_autoclose_checkbox.setToolTip("Auto Close the GUI when pressing start button.")
layout.addWidget(self.use_autoclose_checkbox, 130 + (len(gpu_stats) * 2), 0)
self.listen_port_textfield = QLineEdit()
self.listen_port_textfield.setPlaceholderText("Enter port number")
self.listen_port_textfield.setEnabled(False)
layout.addWidget(self.listen_port_textfield, 140 + (len(gpu_stats) * 2), 1)
self.start_button = QPushButton("Start")
self.start_button.setToolTip("Starts the Webui with the settings set by this GUI")
self.start_button.clicked.connect(self.on_start_button_clicked)
layout.addWidget(self.start_button, 140 + (len(gpu_stats) * 2), 0)
self.save_button = QPushButton("Save Settings")
self.save_button.setToolTip("You can Save your current Settings. Neat, isn't it?")
self.save_button.clicked.connect(self.on_save_button_clicked)
layout.addWidget(self.save_button, 150 + (len(gpu_stats) * 2), 0)
# Load Button
self.load_button = QPushButton("Load")
self.load_button.setToolTip("It's a button. That loads a selected Profile. Sometimes, I'm just create explaining things.")
self.load_button.clicked.connect(self.on_load_button_clicked)
layout.addWidget(self.load_button, 150 + (len(gpu_stats) * 2), 1)
# Show if Update is available
self.update_button_ui = QPushButton("Update\nAvailable")
self.update_button_ui.setToolTip("Shows if an update is available")
self.update_button_ui.setStyleSheet("QPushButton { color: #ff9999; font-weight: bold; }")
self.update_button_ui.clicked.connect(self.on_update_button_ui_clicked)
layout.addWidget(self.update_button_ui, 150 + (len(gpu_stats) * 2), 2, 2, 2)
self.update_button_ui.setVisible(False)
# Textfield for the Profile Name
self.profile_name_textfield = QLineEdit()
self.profile_name_textfield.setPlaceholderText("Enter Name for the Profile, keep empty to overwrite default")
self.profile_name_textfield.setToolTip("You can leave this blank, then only the default profile will be overwritten. If you want to get some organizing done, you can name it. For example:\nProfile for RP\nProfile for Chat\nProfile for coding\nProfile for Superbooga\nERROR: 404 no limits found")
layout.addWidget(self.profile_name_textfield, 151 + (len(gpu_stats) * 2), 0)
# Profiles Dropdown
self.profiles_dropdown = QComboBox()
self.populate_profiles_dropdown()
#self.profiles_dropdown.setPlaceholderText("Choose Profile to load")
self.profiles_dropdown.setToolTip("Here you can choose which profile you want to load. Choose, Load, Profit.")
#layout.addWidget(QLabel("Choose Profile:"), 84 + (len(gpu_stats) * 2), 1)
layout.addWidget(self.profiles_dropdown, 151 + (len(gpu_stats) * 2), 1)
central_widget = QWidget()
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)
def on_api_streaming_port_checkbox_changed(self, state):
if not self.api_public_checkbox.isChecked() and self.api_checkbox.isChecked():
self.api_streaming_port_SpinBox.setEnabled(state == Qt.Checked)
def on_api_blocking_port_checkbox_changed(self, state):
if not self.api_public_checkbox.isChecked() and self.api_checkbox.isChecked():
self.api_blocking_port_SpinBox.setEnabled(state == Qt.Checked)
def on_api_settings_checkbox_stateChanged(self, state):
self.api_header.setVisible(state == Qt.Checked)
self.api_checkbox.setVisible(state == Qt.Checked)
self.api_blocking_port_checkbox.setVisible(state == Qt.Checked)
self.api_blocking_port_SpinBox.setVisible(state == Qt.Checked)
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.toolboxapiline.setVisible(state == Qt.Checked)
def on_rwkv_settings_checkbox_stateChanged(self, state):
self.rwkv_header.setVisible(state == Qt.Checked)
self.rwkv_checkbox.setVisible(state == Qt.Checked)
self.rwkv_strategy_checkbox.setVisible(state == Qt.Checked)
self.rwkv_strategy_dropdown.setVisible(state == Qt.Checked)
self.rwkv_allocation_spinbox.setVisible(state == Qt.Checked)
self.rwkv_cuda_checkbox.setVisible(state == Qt.Checked)
self.rwkv_line.setVisible(state == Qt.Checked)
def on_flexgen_settings_checkbox_stateChanged(self, state):
self.flexgen_header_label.setVisible(state == Qt.Checked)
self.flexgen_checkbox.setVisible(state == Qt.Checked)
self.flexgen_percentage_label.setVisible(state == Qt.Checked)
self.flexgen_percentage_spinbox1.setVisible(state == Qt.Checked)
self.flexgen_percentage_spinbox2.setVisible(state == Qt.Checked)
self.flexgen_percentage_spinbox3.setVisible(state == Qt.Checked)
self.flexgen_percentage_spinbox4.setVisible(state == Qt.Checked)
self.flexgen_percentage_spinbox5.setVisible(state == Qt.Checked)
self.flexgen_percentage_spinbox6.setVisible(state == Qt.Checked)
self.flexgen_compression_checkbox.setVisible(state == Qt.Checked)
self.flexgen_pin_weight_label.setVisible(state == Qt.Checked)
self.flexgen_pin_weight_dropdown.setVisible(state == Qt.Checked)
self.flexline.setVisible(state == Qt.Checked)
def on_llama_settings_checkbox_stateChanged(self, state):
self.llama_label_header.setVisible(state == Qt.Checked)
self.llama_threads_spinbox.setVisible(state == Qt.Checked)
self.llama_threads_label.setVisible(state == Qt.Checked)
self.llama_batch_size_label.setVisible(state == Qt.Checked)
self.llama_batch_size_spinbox.setVisible(state == Qt.Checked)
self.llama_mmap_checkbox.setVisible(state == Qt.Checked)
self.llama_mlock_checkbox.setVisible(state == Qt.Checked)
self.llama_cache_capacity_label.setVisible(state == Qt.Checked)
self.llama_cache_capacity_spinbox.setVisible(state == Qt.Checked)
self.llama_line.setVisible(state == Qt.Checked)
self.llama_gpu_layer_label.setVisible(state == Qt.Checked)
self.llama_gpu_layer_spinbox.setVisible(state == Qt.Checked)
self.llama_cache_capacity_units.setVisible(state == Qt.Checked)
self.llama_n_ctx_label.setVisible(state == Qt.Checked)
self.llama_n_ctx_dropdown.setVisible(state == Qt.Checked)
self.llama_seed_label.setVisible(state == Qt.Checked)
self.llama_seed_spinbox.setVisible(state == Qt.Checked)
def on_deepspeed_nvme_button_clicked(self):
folder = QFileDialog.getExistingDirectory(self, "Offload Directory")
if folder:
self.selected_offload_directory = folder
self.deepspeed_nvme_current_label.setText(f"Current Directory Folder: {self.selected_offload_directory}")
else:
self.selected_offload_directory = none
def on_deepspeed_nvme_checkbox_changed(self, state):
self.deepspeed_nvme_label.setVisible(state == Qt.Checked)
self.deepspeed_nvme_current_label.setVisible(state == Qt.Checked)
self.deepspeed_nvme_button.setVisible(state == Qt.Checked)
def on_deepspeed_settings_checkbox_stateChanged(self, state):
self.deepspeed_label_header.setVisible(state == Qt.Checked)
self.deepspeed_checkbox.setVisible(state == Qt.Checked)
self.deepspeed_local_rank_label.setVisible(state == Qt.Checked)
self.deepspeed_local_rank_spinbox.setVisible(state == Qt.Checked)
self.deepspeed_line.setVisible(state == Qt.Checked)
self.deepspeed_gpu_num_label.setVisible(state == Qt.Checked)
self.deepspeed_gpu_num_spinbox.setVisible(state == Qt.Checked)
self.deepspeed_nvme_checkbox.setVisible(state == Qt.Checked)
def on_update_button_ui_clicked(self):
self.show_version_window()
def update_check(self):
latest_version = self.get_latest_version()
if latest_version and latest_version > version:
self.update_button_ui.setVisible(True)
def show_version_window(self):
latest_version = self.get_latest_version()
if latest_version and latest_version > version:
release_notes = self.get_release_notes()
update_text = f"A new version ({latest_version}) is available! Do you want to update?\n\n\n{release_notes}"
reply = QMessageBox.question(self, "Update Available", update_text, QMessageBox.Yes | QMessageBox.No)
if reply == QMessageBox.Yes:
release_url = f"https://github.com/Pakobbix/StartUI-oobabooga-webui/releases/tag/{latest_version}"
if sys.platform == "win32":
os.startfile(release_url)
else:
try:
subprocess.Popen(["xdg-open", release_url])
except OSError:
self.show_error_message("Error", f"Could not open the link. Please open it manually.\n{release_url}")
def on_report_bug_clicked(self):
github_new_issue = "https://github.com/Pakobbix/StartUI-oobabooga-webui/issues/new"
if sys.platform == "win32":
os.startfile(github_new_issue)
else:
try:
subprocess.Popen(["xdg-open", github_new_issue])
except OSError:
self.show_error_message("Error", f"Could not open the link. Please open it manually.\n{github_new_issue}")
def on_Github_clicked(self):
startui_url = "https://github.com/Pakobbix/StartUI-oobabooga-webui/"
if sys.platform == "win32":
os.startfile(startui_url)
else:
try:
subprocess.Popen(["xdg-open", startui_url])
except OSError:
self.show_error_message("Error", f"Could not open the link. Please open it manually.\n{startui_url}")
def on_oobabooga_clicked(self):
oobabooga_url = "https://github.com/oobabooga/text-generation-webui"
if sys.platform == "win32":
os.startfile(oobabooga_url)
else:
try:
subprocess.Popen(["xdg-open", oobabooga_url])
except OSError:
self.show_error_message("Error", f"Could not open the link. Please open it manually.\n{oobabooga_url}")
def get_latest_version(self):
try:
url = "https://api.github.com/repos/Pakobbix/StartUI-oobabooga-webui/releases/latest"
response = requests.get(url)
if response.status_code == 200:
latest_release = response.json()
tag_name = latest_release["tag_name"]
return tag_name
else:
return None
except Exception as e:
print(f"Error fetching latest version: {str(e)}")
return None
def get_release_notes(self):
try:
url = "https://api.github.com/repos/Pakobbix/StartUI-oobabooga-webui/releases/latest"
response = requests.get(url)
if response.status_code == 200:
latest_release = response.json()
release_notes = latest_release["body"]
return release_notes
else:
return None
except Exception as e:
print(f"Error fetching release notes: {str(e)}")
return None
def show_about_window(self, action):
latest_version = self.get_latest_version()
release_url = f"https://github.com/Pakobbix/StartUI-oobabself.offload_directoryooga-webui/releases/tag/{latest_version}"
if latest_version and latest_version > version:
about_text = f"A new version ({latest_version}) is available! Please update.
StartUI for oobabooga's webui.
Current Version: {version}
This is an GUI (Graphical User Interface), to set flags depending on the user selection."
else:
about_text = f"StartUI for oobabooga's webui.\n\nVersion: {version}\n\nThis is an GUI (Graphical User Interface), to set flags depending on the user selection."
QMessageBox.about(self, "About", about_text)
def on_use_extensions_checkbox_changed(self, state):
self.extensions_list.setVisible(state == Qt.Checked)
def on_use_lora_checkbox_changed(self, state):
self.lora_list.setVisible(state == Qt.Checked)
def on_use_disk_checkbox_changed(self, state):
self.change_disk_cache_checkbox.setVisible(state == Qt.Checked)
self.current_disk_cache_label.setVisible(state == Qt.Checked)
if not self.use_disk_checkbox.isChecked():
self.choose_disk_folder_label.setVisible(False)
self.choose_disk_folder_button.setVisible(False)
if self.use_disk_checkbox.isChecked() and self.change_disk_cache_checkbox.isChecked():
self.choose_disk_folder_label.setVisible(True)
self.choose_disk_folder_button.setVisible(True)
if state == Qt.Checked:
# Check if disk cache path is empty
if not self.disk_cache_path:
# Set default disk cache path
self.disk_cache_path = "/cache"
# Update the label text
self.current_disk_cache_label.setText(f"Current Cache Folder: {self.disk_cache_path}")
def on_change_disk_cache_checkbox_changed(self, state):
self.choose_disk_folder_label.setVisible(state == Qt.Checked)
self.choose_disk_folder_button.setVisible(state == Qt.Checked)
def on_choose_disk_folder_button_clicked(self):
folder = QFileDialog.getExistingDirectory(self, "Choose Disk Cache Folder")
if folder:
self.disk_cache_textfield.setText(folder)
self.disk_cache_path = folder
self.current_disk_cache_label.setText(f"Current Cache Folder: {self.disk_cache_path}")
def show_error_message(self, message):
QMessageBox.critical(self, "Error", message)
def reload_models(self):
model_folders = [name for name in os.listdir(model_folder) if os.path.isdir(os.path.join(model_folder, name))]
self.model_dropdown.clear() # Clear the existing items
self.model_dropdown.addItems(model_folders) # Add the updated items
def set_ram_slider_max(self):
ram_size = psutil.virtual_memory().available
ram_size_gb = ram_size // (1024 ** 3)
self.ram_slider.setMaximum(ram_size_gb)
def on_ram_slider_changed(self, value):
self.ram_value_label.setText(f"{value} GiB")
def on_pre_layer_slider_changed(self, value, idx):
# Calculate the current total value of all sliders
total_value = sum(slider.value() for slider in self.pre_layer_slider)
# Check if the total value exceeds the maximum
if total_value > self.pre_layer_amount_max:
# Calculate the maximum allowed value for the current slider
max_allowed_value = self.pre_layer_amount_max - (total_value - value)
# Adjust the value of the current slider if necessary
if value > max_allowed_value:
self.pre_layer_slider[idx].setValue(max_allowed_value)
value = max_allowed_value
self.pre_layer_slider_value[idx].setText(str(value))
def on_vram_slider_changed(self, value, gpu_idx):
self.gpu_vram_labels[gpu_idx].setText(f"{value} GiB")
#print(f"GPU {gpu_idx} VRAM usage: {value} GiB")
def on_gpu_radio_button_toggled(self, checked):
# Hide/show GPU-related widgets
for slider, label_vram, label_gpu in zip(self.gpu_vram_sliders, self.gpu_vram_labels, self.gpu_labels):
slider.setVisible(checked)
label_vram.setVisible(checked)
label_gpu.setVisible(checked)
# Hide RAM slider and value label
self.ram_label.hide()
self.ram_slider.hide()
self.ram_value_label.hide()
# Uncheck CPU and Autodevice radio buttons
if checked:
self.cpu_radio_button.setChecked(False)
self.auto_radio_button.setChecked(False)
def on_cpu_radio_button_toggled(self, checked):
# Hide/show GPU-related widgets
if nvidia_gpu:
for slider, label_vram, label_gpu in zip(self.gpu_vram_sliders, self.gpu_vram_labels, self.gpu_labels):
slider.hide()
label_vram.hide()
label_gpu.hide()
# Show RAM slider and value label
self.ram_label.setVisible(checked)
self.ram_slider.setVisible(checked)
self.ram_value_label.setVisible(checked)
# Uncheck GPU and Autodevice radio buttons
if checked and nvidia_gpu:
self.gpu_radio_button.setChecked(False)
self.auto_radio_button.setChecked(False)
elif checked and not nvidia_gpu:
self.auto_radio_button.setChecked(False)
def on_auto_radio_button_toggled(self, checked):
# Hide/show GPU-related widgets
if nvidia_gpu:
for slider, label_vram, label_gpu in zip(self.gpu_vram_sliders, self.gpu_vram_labels, self.gpu_labels):
slider.hide()
label_vram.hide()
label_gpu.hide()
# Hide RAM slider and value label
self.ram_label.hide()
self.ram_slider.hide()
self.ram_value_label.hide()
# Uncheck GPU and CPU radio buttons
if checked and nvidia_gpu:
self.gpu_radio_button.setChecked(False)
self.cpu_radio_button.setChecked(False)
elif checked and not nvidia_gpu:
self.cpu_radio_button.setChecked(False)
def on_listen_port_checkbox_changed(self, state):
self.listen_port_textfield.setEnabled(state == Qt.Checked)
def on_authentication_checkbox_changed(self, state):
self.choose_file_label.setVisible(state == Qt.Checked)
self.choose_file_button.setVisible(state == Qt.Checked)
def on_choose_file_button_clicked(self):
file_dialog = QFileDialog(self)
file_dialog.setFileMode(QFileDialog.ExistingFile)
file_dialog.setWindowTitle("Choose Authentication File")
file_dialog.setNameFilter("All Files (*)")
if file_dialog.exec_():
selected_files = file_dialog.selectedFiles()
if selected_files:
file_path = selected_files[0]
self.choose_file_label.setText(file_path)
def on_save_button_clicked(self):
settings = {
"model": self.model_dropdown.currentText(), # Saves the Current selected Model
"model_type": self.model_type.currentText(), # Saves the Current selected Model Type
"wbits": self.wbit_dropdown.currentText(), # Saves the WBIT Setting
"groupsize": self.gsize_dropdown.currentText(), # Saves the Groupsize
"mode": self.mode_dropdown.currentText(), # Saves the selected interaction mode (Chat,cai_chat,Notebook)
"use_gpu": self.gpu_radio_button.isChecked(), # Save the state of the GPU radio button
"use_cpu": self.cpu_radio_button.isChecked(), # Save the state of the CPU radio button
"use_auto": self.auto_radio_button.isChecked(), # Save the state of the auto device radio button
"built_in_ram": self.ram_slider.value(), # Save the value of the built-in RAM slider
"use_8bit": self.use_8bit_checkbox.isChecked(), # Saves the state of the 8bit checkbox
"no_stream": self.use_nostream_checkbox.isChecked(), # Saves the state of the no_stream checkbox
"use_16bit": self.use_16bit_checkbox.isChecked(), # Saves the state of the use_16bit checkbox
"use_disk": self.use_disk_checkbox.isChecked(), # Saves the state of the use_disk checkbox
"disk_cache": self.disk_cache_textfield.text(), # Saves the state of the disk_cache textfield
"xformers": self.use_xformers_checkbox.isChecked(), # Saves the state of the xformers checkbox
"trust_remote_code": self.use_trc_checkbox.isChecked(), # Saves the state of the trust_remote_code checkbox
"monkeypatch": self.use_monkey_checkbox.isChecked(), # Saves the state of the monkeypatch checkbox
"quant_attn": self.use_quant_checkbox.isChecked(), # Saves the state of the quant_attn checkbox
"multimodal": self.use_multimodal_checkbox.isChecked(), # Saves the state of the multimodal checkbox
"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
"deepspeed_nvme_enabled": self.deepspeed_nvme_checkbox.isChecked(), # Saves the state of the deepspeed_nvme_checkbox
"deepspeed_nvme_path": self.selected_offload_directory, # Saves the state of the offload_directory
"deepspeed_local_rank": self.deepspeed_local_rank_spinbox.value(), # Saves the state of the deepspeed_local_rank_spinbox
"llama_settings": self.llama_settings_checkbox.isChecked(), # Saves the state of the llama_settings_checkbox
"llama_threads": self.llama_threads_spinbox.value(), # Saves the state of the llama_threads_spinbox
"llama_batch_size": self.llama_batch_size_spinbox.value(), # Saves the state of the llama_batch_size_spinbox
"llama_no_map": self.llama_mmap_checkbox.isChecked(), # Saves the state of the llama_no_map_checkbox
"llama_use_mlock": self.llama_mlock_checkbox.isChecked(), # Saves the state of the llama_mlock_checkbox
"llama_cache_capacity": self.llama_cache_capacity_spinbox.value(), # Saves the state of the llama_cache_capacity_spinbox
"llama_cache_units": self.llama_cache_capacity_units.currentText(), # Saves the state of the llama_cache_capacity_units
"llama_gpu_layer": self.llama_gpu_layer_spinbox.value(), # Saves the state of the llama_gpu_layer_spinbox
"llama_n_ctx": self.llama_n_ctx_dropdown.currentText(), # Saves the state of the llama_n_ctx_dropdown
"llama_seed": self.llama_seed_spinbox.value(), # Saves the state of the llama_seed_spinbox
"flexgen_settings": self.flexgen_settings_checkbox.isChecked(), # Saves the state of the flexgen_settings_checkbox
"use_flexgen": self.flexgen_checkbox.isChecked(), # Saves the state of the flexgen_checkbox
"flexgen_precentage_1": self.flexgen_percentage_spinbox1.value(), # Saves the state of the flexgen_percentage_spinbox1
"flexgen_precentage_2": self.flexgen_percentage_spinbox2.value(), # Saves the state of the flexgen_percentage_spinbox2
"flexgen_precentage_3": self.flexgen_percentage_spinbox3.value(), # Saves the state of the flexgen_percentage_spinbox3
"flexgen_precentage_4": self.flexgen_percentage_spinbox4.value(), # Saves the state of the flexgen_percentage_spinbox4
"flexgen_precentage_5": self.flexgen_percentage_spinbox5.value(), # Saves the state of the flexgen_percentage_spinbox5
"flexgen_precentage_6": self.flexgen_percentage_spinbox6.value(), # Saves the state of the flexgen_percentage_spinbox6
"flexgen_compression": self.flexgen_compression_checkbox.isChecked(), # Saves the state of the flexgen_compression_checkbox
"flexgen_pin_weight": self.flexgen_pin_weight_dropdown.currentText(), # Saves the state of the flexgen_pin_weight_dropdown
"rwkv_settings": self.rwkv_settings_checkbox.isChecked(), # Saves the state of the rwkv_settings_checkbox
"use_rwkv": self.rwkv_checkbox.isChecked(), # Saves the state of the rwkv_checkbox
"rwkv_strategy": self.rwkv_strategy_checkbox.isChecked(), # Saves the state of the rwkv_strategy_checkbox
"rwkv_strategy_dropdown": self.rwkv_strategy_dropdown.currentText(), # Saves the state of the rwkv_strategy_dropdown
"rwkv_allocation": self.rwkv_allocation_spinbox.value(), # Saves the state of the rwkv_allocation_spinbox
"rwkv_cuda": self.rwkv_cuda_checkbox.isChecked(), # Saves the state of the rwkv_cuda_checkbox
"api_settings": self.api_settings_checkbox.isChecked(), # Saves the state of the api_settings_checkbox
"use_api": self.api_checkbox.isChecked(), # Saves the state of the api_checkbox
"api_blocking_port_enabled": self.api_blocking_port_checkbox.isChecked(), # Saves the state of the api_blocking_port_checkbox
"api_blocking_port": self.api_blocking_port_SpinBox.value(), # Saves the state of the api_blocking_port_SpinBox
"api_streaming_port_enabled": self.api_streaming_port_checkbox.isChecked(), # Saves the state of the api_streaming_port_checkbox
"api_streaming_port": self.api_streaming_port_SpinBox.value(), # Saves the state of the api_streaming_port_SpinBox
"public_api": self.api_public_checkbox.isChecked(), # Saves the state of the api_public_checkbox
"autotune": self.use_autotune_checkbox.isChecked(), # Saves the state of the autotune checkbox
"autolaunch": self.use_autolaunch_checkbox.isChecked(), # Saves the state of the autotune checkbox
"autoclose": self.use_autoclose_checkbox.isChecked(), # Saves the state of the autotune checkbox
"nocache": self.use_nocache_checkbox.isChecked(), # Saves the state of the autotune checkbox
"listen": self.use_network_checkbox.isChecked(), # Saves the state of the Local Network Checkbox
"listen_port": self.listen_port_checkbox.isChecked(), # Saves the state of the Listen Port Checkbox
"port_number": self.listen_port_textfield.text(), # Saves the Port given in the Textbox
"authentication": self.authentication_checkbox.isChecked(), # Saves the state of the Authentication
"authentication_file": self.choose_file_label.text(), # Save the authentication file path
"character": self.character_to_load.currentText(), # Saves the Characters given in the Textbox
"use_extension": self.use_extensions_checkbox.isChecked(), # Saves the state of the Extension Checkbox
"extensions": [self.extensions_list.item(i).text() for i in range(self.extensions_list.count()) if self.extensions_list.item(i).checkState() == Qt.Checked], # Saves the chosen Extensions
"use_lora": self.use_lora_checkbox.isChecked(), # Saves the state of the Lora Checkbox
"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]
# Get the text entered in the text field
profile_name = self.profile_name_textfield.text()
if not profile_name:
profile_name = "default"
file_path = os.path.join(profiles_folder, f"{profile_name}.json")
with open(file_path, "w") as file:
json.dump(settings, file, indent=4)
def expression_check(self, command):
selected_model = self.model_dropdown.currentText()
# Use a regular expression to check if the selected model matches the pattern
if re.search(r".*mpt.*7b", selected_model, re.IGNORECASE):
# Run the additional commands
run_cmd_with_conda("pip install einops && exit")
elif re.search(r".*vicuna.*7b", selected_model, re.IGNORECASE):
pass
def on_start_button_clicked(self):
command = ""
# LLama Stuff
# llama.cpp threads
if self.llama_threads_spinbox.value() != 0:
command += f" --threads {self.llama_threads_spinbox.value()}"
command += f" --n_batch {self.llama_batch_size_spinbox.value()}"
command += f" --cache-capacity {self.llama_cache_capacity_spinbox.value()}{self.llama_cache_capacity_units.currentText()}"
command += f" --n_ctx {self.llama_n_ctx_dropdown.currentText()}"
command += f" --llama_cpp_seed {self.llama_seed_spinbox.value()}"
if self.llama_gpu_layer_spinbox.value() != 0:
command += f" --n-gpu-layers {self.llama_gpu_layer_spinbox.value()}"
if self.llama_mmap_checkbox.isChecked():
command += " --no-map"
if self.llama_mlock_checkbox.isChecked():
command += " --mlock"
# FlexGen Commands
if self.flexgen_checkbox.isChecked():
command += " --flexgen"
command += f" --percent {self.flexgen_percentage_spinbox1.value()} {self.flexgen_percentage_spinbox2.value()} {self.flexgen_percentage_spinbox3.value()} {self.flexgen_percentage_spinbox4.value()} {self.flexgen_percentage_spinbox5.value()} {self.flexgen_percentage_spinbox6.value()}"
if self.flexgen_compression_checkbox.isChecked():
command += " --compression-weight"
if self.flexgen_pin_weight_dropdown.currentText() != "none":
command += f" --pin-weight {self.flexgen_pin_weight_dropdown.currentText()}"
# Add the chosen model to the command
chosen_model = self.model_dropdown.currentText()
if self.model_dropdown.currentText() != "none":
command += f" --model {chosen_model}"
# Add the chosen model type to the command
if self.model_type.currentText() != "none" and self.model_dropdown.currentText() != "none":
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]
if self.use_lora_checkbox.isChecked() and self.model_dropdown.currentText() != "none":
if loras:
command += f" --lora {' '.join(loras)}"
# Add Characters to the command
if self.character_to_load.currentText() != "none":
command += f" --character {self.character_to_load.currentText()}"
# Adds wbits to the command, if not "none"
if self.wbit_dropdown.currentText() != "none":
if not self.cpu_radio_button.isChecked() and self.model_dropdown.currentText() != "none":
command += f" --wbits {self.wbit_dropdown.currentText()}"
# Adds Groupsize to the command, if not "none"
if self.gsize_dropdown.currentText() != "none":
if not self.cpu_radio_button.isChecked() and self.model_dropdown.currentText() != "none":
command += f" --groupsize {self.gsize_dropdown.currentText()}"
# Add the chosen mode to the command (Chat, cai-chat, notebook)
chosen_mode = self.mode_dropdown.currentText()
command += f" --{chosen_mode}"
# Handle GPU or CPU selection
if self.gpu_radio_button.isChecked():
# GPU radio button is selected
total_vram = sum(slider.value() for slider in self.gpu_vram_sliders)
if total_vram == 0:
error_message = "Error:\nAt least one VRAM value must be greater than 0 for GPU execution."
self.show_error_message(error_message)
else:
command += " --gpu-memory"
for vram in [slider.value() for slider in self.gpu_vram_sliders]:
if vram > 0:
command += f" {vram}"
elif self.cpu_radio_button.isChecked():
# CPU radio button is selected
command += " --cpu-memory"
ram_value = self.ram_slider.value()
if ram_value > 0:
command += f" {ram_value}"
else:
# Display an error message in a dialog box when RAM value is 0
error_message = "Error:\nRAM value cannot be 0 for CPU execution."
self.show_error_message(error_message)
return
# Auto Device is Activated:
elif self.auto_radio_button.isChecked():
command += " --auto-device"
# Add 8bit loading
if self.use_8bit_checkbox.isChecked():
command += " --load-in-8bit"
# Add no-stream
if self.use_nostream_checkbox.isChecked():
command += " --no-stream"
# Add 16bit full precision loading
if self.use_16bit_checkbox.isChecked():
command += " --bf16"
# Add xformers loading
if self.use_xformers_checkbox.isChecked():
command += " --xformers"
# Use "Trust Remote Code=TRUE" for ex. MPT-7B
if self.use_trc_checkbox.isChecked():
command += " --trust-remote-code"
if re.search(r"mpt.*7b", chosen_model):
if not self.use_trc_checkbox.isChecked():
command += " --trust-remote-code"
# Use Triton Warmup & Autotune
if self.use_autotune_checkbox.isChecked():
command += " --warmup_autotune"
# Add loading with Monkey Patch
if self.use_monkey_checkbox.isChecked():
command += " --monkey-patch"
# Enable quant attention
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"
# Add --auto-launch
if self.use_autolaunch_checkbox.isChecked():
command += " --auto-launch"
# Local Network Mode
if self.use_network_checkbox.isChecked():
command += " --listen"
# Multimodal Mode
if self.use_multimodal_checkbox.isChecked():
command += " --multimodal-pipeline"
# Use Disk to store part of the Model
if self.use_disk_checkbox.isChecked():
command += " --disk"
if self.change_disk_cache_checkbox.isChecked():
if self.disk_cache_textfield.text():
command += f" --disk-cache-dir {self.disk_cache_textfield.text()}"
# Add listen port if the checkbox is checked and a port number is provided
if self.listen_port_checkbox.isChecked():
listen_port = self.listen_port_textfield.text()
if listen_port.isdigit():
command += f" --listen-port {listen_port}"
# Adds the authentication to the command, if active
if self.authentication_checkbox.isChecked():
if self.choose_file_label.text():
command += f" --gradio-auth-path {self.choose_file_label.text()}"
## Adds the Prelayer selection
slider_values = [slider.value() for slider in self.pre_layer_slider]
if any(value > 0 for value in slider_values):
command += f" --pre_layer {' '.join(str(value) for value in slider_values if value > 0)}"
# IF sdp_attention is checked
if self.use_sdp_attention_checkbox.isChecked():
command += " --sdp-attention"
# If AutoGPTQ is checked
if self.use_autogptq_checkbox.isChecked():
command += " --autogptq"
# If triton is checked
if self.use_triton_checkbox.isChecked():
command += " --triton"
# Adds the chosen extensions to the list of the command.
extensions = [self.extensions_list.item(i).text() for i in range(self.extensions_list.count()) if self.extensions_list.item(i).checkState() == Qt.Checked]
if self.use_extensions_checkbox.isChecked():
if extensions:
command += f" --extensions {' '.join(extensions)}"
if self.api_checkbox.isChecked():
command += " --api"
if self.api_public_checkbox.isChecked():
command += " --public-api"
if self.api_checkbox.isChecked() and not self.api_public_checkbox.isChecked():
if self.api_blocking_port_checkbox.isChecked():
command += f" --api-blocking-port {self.api_blocking_port_SpinBox.text()}"
if self.api_streaming_port_checkbox.isChecked():
command += f" --api-streaming-port {self.api_streaming_port_SpinBox.text()}"
# Just for debugging.
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
if self.deepspeed_checkbox.isChecked():
if platform.system() == "Linux":
gpu_number = self.deepspeed_gpu_num_spinbox.text()
deepspeed_command = f"deepspeed --num_gpus={gpu_number} ./text-generation-webui/server.py --deepspeed"
if self.deepspeed_nvme_checkbox.isChecked():
deepspeed_command += f" --nvme-offload-dir {self.offload_directory}"
if self.deepspeed_local_rank_spinbox.text() != "0":
deepspeed_command += f" --local_rank {self.deepspeed_local_rank_spinbox.text()}"
run_cmd_with_conda(f"pip install deepspeed ; clear && {deepspeed_command} {command}")
elif platform.system() == "Windows":
message = "DeepSpeed is currently not supported on Windows"
QMessageBox.critical(self, "Error", message)
if not self.deepspeed_checkbox.isChecked():
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(f"python {webui_file} --update && exit")
def populate_profiles_dropdown(self):
self.profiles_dropdown.clear()
profiles = [name for name in os.listdir(profiles_folder) if name.endswith(".json")]
self.profiles_dropdown.addItems(profiles)
def on_load_button_clicked(self):
selected_profile = self.profiles_dropdown.currentText()
profile_file = os.path.join(profiles_folder, selected_profile)
self.load_profile(profile_file)
# Populate the profile name text field with the loaded profile name
profile_name = selected_profile.replace(".json", "")
self.profile_name_textfield.setText(profile_name)
def apply_load_settings(self, settings):
self.model_dropdown.setCurrentText(settings.get("model", ""))
self.model_type.setCurrentText(settings.get("model_type", ""))
self.wbit_dropdown.setCurrentText(settings.get("wbits", ""))
self.gsize_dropdown.setCurrentText(settings.get("groupsize", ""))
self.mode_dropdown.setCurrentText(settings.get("mode", ""))
use_gpu = settings.get("use_gpu", False)
use_cpu = settings.get("use_cpu", False)
self.gpu_radio_button.setChecked(use_gpu)
self.cpu_radio_button.setChecked(use_cpu)
built_in_ram = settings.get("built_in_ram", 0)
self.ram_slider.setValue(built_in_ram)
self.listen_port_checkbox.setChecked(settings.get("listen_port", False))
self.listen_port_textfield.setText(settings.get("port_number", ""))
self.use_8bit_checkbox.setChecked(settings.get("use_8bit", False))
self.use_nostream_checkbox.setChecked(settings.get("no_stream", False))
self.use_16bit_checkbox.setChecked(settings.get("use_16bit", False))
self.use_disk_checkbox.setChecked(settings.get("use_disk", False))
self.disk_cache_textfield.setText(settings.get("disk_cache", ""))
self.current_disk_cache_label.setText(f"Current folder: {settings.get('disk_cache', '')}")
self.use_xformers_checkbox.setChecked(settings.get("xformers", False))
self.use_trc_checkbox.setChecked(settings.get("trust_remote_code", False))
self.use_monkey_checkbox.setChecked(settings.get("monkeypatch", False))
self.use_quant_checkbox.setChecked(settings.get("quant_attn", False))
self.use_multimodal_checkbox.setChecked(settings.get("multimodal", False))
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))
# Acceleration 4bit
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))
# Deepspeed
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)))
self.selected_offload_directory = settings.get("deepspeed_nvme_path", "")
self.deepspeed_nvme_current_label.setText(f"Current Directory Folder: {self.selected_offload_directory}")
self.deepspeed_nvme_checkbox.setChecked(settings.get("deepspeed_nvme_enabled", False))
self.deepspeed_local_rank_spinbox.setValue(int(settings.get("deepspeed_local_rank", 0)))
# llama
self.llama_settings_checkbox.setChecked(settings.get("llama_settings", False))
self.llama_threads_spinbox.setValue(int(settings.get("llama_threads", 0)))
self.llama_batch_size_spinbox.setValue(int(settings.get("llama_batch_size", 0)))
self.llama_mmap_checkbox.setChecked(settings.get("llama_no_map", False))
self.llama_mlock_checkbox.setChecked(settings.get("llama_use_mlock", False))
self.llama_cache_capacity_spinbox.setValue(int(settings.get("llama_cache_capacity", 0)))
self.llama_cache_capacity_units.setCurrentText(settings.get("llama_cache_units", ""))
self.llama_gpu_layer_spinbox.setValue(int(settings.get("llama_gpu_layer", 0)))
self.llama_n_ctx_dropdown.setCurrentText(settings.get("llama_n_ctx", ""))
self.llama_seed_spinbox.setValue(int(settings.get("llama_seed", 0)))
# flexgen
self.flexgen_settings_checkbox.setChecked(settings.get("flexgen_settings", False))
self.flexgen_checkbox.setChecked(settings.get("use_flexgen", False))
self.flexgen_percentage_spinbox1.setValue(int(settings.get("flexgen_precentage_1", 0)))
self.flexgen_percentage_spinbox2.setValue(int(settings.get("flexgen_precentage_2", 100)))
self.flexgen_percentage_spinbox3.setValue(int(settings.get("flexgen_precentage_3", 100)))
self.flexgen_percentage_spinbox4.setValue(int(settings.get("flexgen_precentage_4", 0)))
self.flexgen_percentage_spinbox5.setValue(int(settings.get("flexgen_precentage_5", 100)))
self.flexgen_percentage_spinbox6.setValue(int(settings.get("flexgen_precentage_6", 0)))
self.flexgen_compression_checkbox.setChecked(settings.get("flexgen_compression", False))
self.flexgen_pin_weight_dropdown.setCurrentText(settings.get("flexgen_pin_weight", ""))
# RWKV
self.rwkv_settings_checkbox.setChecked(settings.get("rwkv_settings", False))
self.rwkv_checkbox.setChecked(settings.get("use_rwkv", False))
self.rwkv_strategy_checkbox.setChecked(settings.get("rwkv_strategy", False))
self.rwkv_strategy_dropdown.setCurrentText(settings.get("rwkv_strategy_dropdown", ""))
self.rwkv_allocation_spinbox.setValue(int(settings.get("rwkv_allocation", 0)))
self.rwkv_cuda_checkbox.setChecked(settings.get("rwkv_cuda", False))
# API
self.api_settings_checkbox.setChecked(settings.get("api_settings", False))
self.api_checkbox.setChecked(settings.get("use_api", False))
self.api_blocking_port_checkbox.setChecked(settings.get("api_blocking_port_enabled", False))
self.api_blocking_port_SpinBox.setValue(int(settings.get("api_blocking_port", 5000)))
self.api_streaming_port_checkbox.setChecked(settings.get("api_streaming_port_enabled", False))
self.api_streaming_port_SpinBox.setValue(int(settings.get("api_streaming_port", 5005)))
self.api_public_checkbox.setChecked(settings.get("public_api", False))
self.use_autotune_checkbox.setChecked(settings.get("autotune", False))
self.use_autolaunch_checkbox.setChecked(settings.get("autolaunch", False))
self.use_autoclose_checkbox.setChecked(settings.get("autoclose", False))
self.use_nocache_checkbox.setChecked(settings.get("nocache", False))
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 "prelayer" in settings:
pre_layer_values = settings["prelayer"]
for i, value in enumerate(pre_layer_values):
self.pre_layer_slider[i].setValue(value)
if nvidia_gpu:
gpu_vram_settings = settings.get("gpu_vram", [])
for idx, slider in enumerate(self.gpu_vram_sliders):
if idx < len(gpu_vram_settings):
slider.setValue(gpu_vram_settings[idx])
self.use_extensions_checkbox.setChecked(settings.get("use_extension", False))
extensions_settings = settings.get("extensions", [])
for i in range(self.extensions_list.count()):
extension = self.extensions_list.item(i).text()
if extension in extensions_settings:
self.extensions_list.item(i).setCheckState(Qt.Checked)
else:
self.extensions_list.item(i).setCheckState(Qt.Unchecked)
self.use_lora_checkbox.setChecked(settings.get("use_lora", False))
lora_settings = settings.get("loras", [])
for i in range(self.lora_list.count()):
lora = self.lora_list.item(i).text()
if lora in lora_settings:
self.lora_list.item(i).setCheckState(Qt.Checked)
else:
self.lora_list.item(i).setCheckState(Qt.Unchecked)
def load_settings(self):
default_profile = os.path.join(profiles_folder, "default.json")
if os.path.exists(default_profile):
with open(default_profile, "r") as file:
try:
settings = json.load(file)
self.apply_load_settings(settings)
except json.JSONDecodeError:
# Handle the case when the file is empty or not in valid JSON format
pass
def load_profile(self, profile_file):
with open(profile_file, "r") as file:
try:
settings = json.load(file)
self.apply_load_settings(settings)
except json.JSONDecodeError:
# Handle the case when the file is empty or not in valid JSON format
pass
if __name__ == "__main__":
app = QApplication(sys.argv)
main_window = MainWindow()
main_window.show()
if darkdetect.isDark():
dark_stylesheet = qdarkstyle.load_stylesheet_pyqt5()
app.setStyleSheet(dark_stylesheet)
sys.exit(app.exec_())