Upgraded the hardware management

api/__init__.py

@@ -1574,7 +1574,7 @@ def prepare_query(self, client_id: str, message_id: int = -1, is_continue: bool
                     self.warning("Couldn't add long term memory information to the context. Please verify the vector database")        # Add information about the user
         user_description=""
         if self.config.use_user_name_in_discussions:
-            user_description="!@>User description:\n"+self.config.user_description
+            user_description="!@>User description:\n"+self.config.user_description+"\n"
 
 
         # Tokenize the conditionning text and calculate its number of tokens
@@ -1971,6 +1971,13 @@ def process_chunk(
                 dt=1
             spd = self.nb_received_tokens/dt
             ASCIIColors.green(f"Received {self.nb_received_tokens} tokens (speed: {spd:.2f}t/s)              ",end="\r",flush=True) 
+            antiprompt = self.personality.detect_antiprompt(self.connections[client_id]["generated_text"])
+            if antiprompt:
+                ASCIIColors.warning(f"\nDetected hallucination with antiprompt: {antiprompt}")
+                self.connections[client_id]["generated_text"] = self.remove_text_from_string(self.connections[client_id]["generated_text"],antiprompt)
+                self.update_message(client_id, self.connections[client_id]["generated_text"], parameters, metadata, None, MSG_TYPE.MSG_TYPE_FULL)
+                return False
+
             self.update_message(client_id, chunk,  parameters, metadata, ui=None, msg_type=message_type)
             return True
         # Stream the generated text to the frontend
@@ -1983,6 +1990,7 @@ def generate(self, full_prompt, prompt, n_predict, client_id, callback=None):
         if self.personality.processor is not None:
             ASCIIColors.info("Running workflow")
             try:
+                self.personality.callback = callback
                 self.personality.processor.run_workflow( prompt, full_prompt, callback)
             except Exception as ex:
                 trace_exception(ex)

app.py

@@ -1254,7 +1254,7 @@ def upgrade_to_gpu(self):
             ASCIIColors.green("PyTorch uninstalled successfully")
             reinstall_pytorch_with_cuda()
             ASCIIColors.yellow("Installing pytorch with cuda support")
-            self.config.enable_gpu=True
+            self.config.hardware_mode="nvidia-tensorcores"
             return jsonify({'status':res==0})
 
 
@@ -1992,7 +1992,7 @@ def clear_personality_files_list(self):
             return jsonify({"state":True})
 
         def start_training(self):
-            if self.config.enable_gpu:
+            if self.config.hardware_mode=="nvidia-tensorcores" or self.config.hardware_mode=="nvidia" or self.config.hardware_mode=="apple-intel" or self.config.hardware_mode=="apple-silicon":
                 if not self.lollms_paths.gptqlora_path.exists():
                     # Clone the repository to the target path
                     ASCIIColors.info("No gptqlora found in your personal space.\nCloning the gptqlora repo")
@@ -2748,17 +2748,6 @@ def sync_cfg(default_config, config):
         if not user_avatar_path.exists():
             # If the user avatar doesn't exist, copy the default avatar from the assets folder
             shutil.copy(default_user_avatar, user_avatar_path)
-        # executor = ThreadPoolExecutor(max_workers=1)
-        # app.config['executor'] = executor
-        # Check if .no_gpu file exists
-        no_gpu_file = Path('.no_gpu')
-        if no_gpu_file.exists():
-            # If the file exists, change self.config.use_gpu to False
-            config.enable_gpu = False
-            config.save_config()
-
-            # Remove the .no_gpu file
-            no_gpu_file.unlink()
 
         bot = LoLLMsWebUI(args, app, socketio, config, config.file_path, lollms_paths)
 

configs/config.yaml

@@ -1,5 +1,5 @@
 # =================== Lord Of Large Language Models Configuration file =========================== 
-version: 39
+version: 40
 binding_name: null
 model_name: null
 
@@ -44,8 +44,8 @@ debug: False
 auto_update: true
 auto_save: true
 auto_title: false
-# Enables gpu usage
-enable_gpu: true
+# Install mode (cpu, cpu-noavx, nvidia-tensorcores, nvidia, amd-noavx, amd, apple-intel, apple-silicon)
+hardware_mode: nvidia-tensorcores
 # Automatically open the browser
 auto_show_browser: true
 

lollms_webui.py

@@ -6,65 +6,36 @@
 This class provides a singleton instance of the LoLLMS web UI, allowing access to its functionality and data across multiple endpoints.
 """
 
-from lollms.app import LollmsApplication
+from lollms.server.elf_server import LOLLMSElfServer
 from lollms.main_config import LOLLMSConfig
 from lollms.paths import LollmsPaths
 
-class LoLLMSWebUI(LollmsApplication):
-    __instance = None
-
-    @staticmethod
-    def build_instance(
-        config: LOLLMSConfig,
-        lollms_paths: LollmsPaths,
-        load_binding=True,
-        load_model=True,
-        try_select_binding=False,
-        try_select_model=False,
-        callback=None,
-        socketio = None
-    ):
-        if LoLLMSWebUI.__instance is None:
-            LoLLMSWebUI(
-                config,
-                lollms_paths,
-                load_binding=load_binding,
-                load_model=load_model,
-                try_select_binding=try_select_binding,
-                try_select_model=try_select_model,
-                callback=callback,
-                socketio=socketio
-            )
-        return LoLLMSWebUI.__instance
-    @staticmethod
-    def get_instance():
-        return LoLLMSWebUI.__instance
-
+class LOLLMSWebUI(LOLLMSElfServer):
     def __init__(
         self,
         config: LOLLMSConfig,
         lollms_paths: LollmsPaths,
         load_binding=True,
         load_model=True,
+        load_voice_service=True,
+        load_sd_service=True,
         try_select_binding=False,
         try_select_model=False,
         callback=None,
         socketio=None
     ) -> None:
         super().__init__(
-            "LoLLMSWebUI",
             config,
             lollms_paths,
             load_binding=load_binding,
             load_model=load_model,
+            load_sd_service=load_sd_service,
+            load_voice_service=load_voice_service,
             try_select_binding=try_select_binding,
             try_select_model=try_select_model,
             callback=callback,
             socketio=socketio
         )
-        if LoLLMSWebUI.__instance is not None:
-            raise Exception("This class is a singleton!")
-        else:
-            LoLLMSWebUI.__instance = self
+        self.app_name = "LOLLMSWebUI"
 
     # Other methods and properties of the LoLLMSWebUI singleton class

new_app.py

@@ -11,7 +11,7 @@
 from lollms.app import LollmsApplication
 from lollms.paths import LollmsPaths
 from lollms.main_config import LOLLMSConfig
-from lollms_webui import LoLLMSWebUI
+from lollms_webui import LOLLMSWebUI
 from pathlib import Path
 from ascii_colors import ASCIIColors
 import socketio
@@ -43,7 +43,9 @@
     if args.port:
         config.port=args.port
 
-    LoLLMSWebUI.build_instance(config=config, lollms_paths=lollms_paths, socketio=sio)
-    from lollms.server.endpoints.lollms_infos import *
+    LOLLMSWebUI.build_instance(config=config, lollms_paths=lollms_paths, socketio=sio)
+
+    # Import all endpoints
+    from lollms.server.endpoints.lollms_infos import router
 
     uvicorn.run(app, host=config.host, port=config.port)

scripts/linux/linux_install.sh

@@ -3,6 +3,22 @@
 # This script will install miniconda and git with all dependencies for this project
 # This enables a user to install this project without manually installing conda and git.
 
+echo "      ___       ___           ___       ___       ___           ___      "
+echo "     /\__\     /\  \         /\__\     /\__\     /\__\         /\  \     "
+echo "    /:/  /    /::\  \       /:/  /    /:/  /    /::|  |       /::\  \    "
+echo "   /:/  /    /:/\:\  \     /:/  /    /:/  /    /:|:|  |      /:/\ \  \   "
+echo "  /:/  /    /:/  \:\  \   /:/  /    /:/  /    /:/|:|__|__   _\:\~\ \  \  "
+echo " /:/__/    /:/__/ \:\__\ /:/__/    /:/__/    /:/ |::::\__\ /\ \:\ \ \__\ "
+echo " \:\  \    \:\  \ /:/  / \:\  \    \:\  \    \/__/~~/:/  / \:\ \:\ \/__/ "
+echo "  \:\  \    \:\  /:/  /   \:\  \    \:\  \         /:/  /   \:\ \:\__\   "
+echo "   \:\  \    \:\/:/  /     \:\  \    \:\  \       /:/  /     \:\/:/  /   "
+echo "    \:\__\    \::/  /       \:\__\    \:\__\     /:/  /       \::/  /    "
+echo "     \/__/     \/__/         \/__/     \/__/     \/__/         \/__/     "
+echo "V8.5 (alpha)"
+echo "-----------------"
+echo "By ParisNeo"
+echo "-----------------"
+
 
 cd "$(dirname "$0")"
 
@@ -19,47 +35,12 @@ if [[ "$PWD" =~ [^#\$\%\&\(\)\*\+\] ]]; then
 fi
 
 
-
-
+export PACKAGES_TO_INSTALL=python=3.11 git
 read -rp "Press Enter to continue..."
 
 clear
 
-echo "      ___       ___           ___       ___       ___           ___      "
-echo "     /\__\     /\  \         /\__\     /\__\     /\__\         /\  \     "
-echo "    /:/  /    /::\  \       /:/  /    /:/  /    /::|  |       /::\  \    "
-echo "   /:/  /    /:/\:\  \     /:/  /    /:/  /    /:|:|  |      /:/\ \  \   "
-echo "  /:/  /    /:/  \:\  \   /:/  /    /:/  /    /:/|:|__|__   _\:\~\ \  \  "
-echo " /:/__/    /:/__/ \:\__\ /:/__/    /:/__/    /:/ |::::\__\ /\ \:\ \ \__\ "
-echo " \:\  \    \:\  \ /:/  / \:\  \    \:\  \    \/__/~~/:/  / \:\ \:\ \/__/ "
-echo "  \:\  \    \:\  /:/  /   \:\  \    \:\  \         /:/  /   \:\ \:\__\   "
-echo "   \:\  \    \:\/:/  /     \:\  \    \:\  \       /:/  /     \:\/:/  /   "
-echo "    \:\__\    \::/  /       \:\__\    \:\__\     /:/  /       \::/  /    "
-echo "     \/__/     \/__/         \/__/     \/__/     \/__/         \/__/     "
-echo " By ParisNeo"
-
-echo "Please specify if you want to use a GPU or CPU."
-echo "*Note* that only NVidea GPUs (cuda) or AMD GPUs (rocm) are supported."
-echo "A) Enable Cuda (for nvidia GPUS)"
-echo "B) Enable ROCm (for AMD GPUs)"
-echo "C) Run CPU mode"
-echo
-read -rp "Input> " gpuchoice
-gpuchoice="${gpuchoice:0:1}"
-
-if [[ "${gpuchoice^^}" == "A" ]]; then
-  PACKAGES_TO_INSTALL="python=3.10 cuda-toolkit ninja git gcc"
-  CHANNEL="-c nvidia/label/cuda-12.1.1 -c nvidia -c conda-forge"
-elif [[ "${gpuchoice^^}" == "B" ]]; then
-  PACKAGES_TO_INSTALL="python=3.10  rocm-comgr rocm-smi ninja git gcc"
-  CHANNEL=" -c conda-forge"
-elif [[ "${gpuchoice^^}" == "C" ]]; then
-  PACKAGES_TO_INSTALL="python=3.10 ninja git gcc"
-  CHANNEL="-c conda-forge"
-else
-  echo "Invalid choice. Exiting..."
-  exit 1
-fi
+
 
 # Better isolation for virtual environment
 unset CONDA_SHLVL
@@ -115,33 +96,28 @@ export CUDA_PATH="$INSTALL_ENV_DIR"
 if [ -d "lollms-webui" ]; then
   cd lollms-webui || exit 1
   git pull
+  git submodule update --init --recursive
+  cd
+  cd lollms-core 
+  pip install -e .
+  cd ..
+  cd utilities\safe_store
+  pip install -e .
+  cd ..\..
+
 else
-  git clone "$REPO_URL"
+  git clone --depth 1  --recurse-submodules "$REPO_URL"
+  git submodule update --init --recursive
+  cd lollms-webui\lollms_core
+  pip install -e .
+  cd ..
+  cd utilities\safe_store
+  pip install -e .
+  cd ..\..
+
   cd lollms-webui || exit 1
 fi
 
-# Initilize all submodules and set them to main branch
-echo "Initializing submodules"
-git submodule update --init
-cd zoos/bindings_zoo
-git checkout main
-cd ../personalities_zoo
-git checkout main
-cd ../extensions_zoo
-git checkout main
-cd ../models_zoo
-git checkout main
-
-cd ../..
-
-cd lollms_core
-git checkout main
-
-cd ../utilities/safe_store
-git checkout main
-
-cd ../..
-
 # Loop through each "git+" requirement and uninstall it (workaround for inconsistent git package updating)
 while IFS= read -r requirement; do
   if echo "$requirement" | grep -q "git+"; then
@@ -152,8 +128,6 @@ done < requirements.txt
 
 # Install the pip requirements
 python -m pip install -r requirements.txt --upgrade
-python -m pip install -e lollms_core --upgrade
-python -m pip install -e utilities/safe_store --upgrade
 
 
 if [[ -e "../linux_run.sh" ]]; then
@@ -180,12 +154,10 @@ else
     cp scripts/linux/linux_update_models.sh ../
 fi
 
-if [[ "${gpuchoice^^}" == "C" ]]; then
-    echo "This is a .no_gpu file." > .no_gpu
-    echo "You have chosen to use only CPU on this system."
-else
-    echo "You have chosen to use GPU on this system."
-fi
+
+cd scripts/python/lollms_installer
+python main.py
+cd ..
 
 PrintBigMessage() {
   echo