Fix api extension

.dockerignore

@@ -1,10 +0,0 @@
-.env
-Dockerfile
-/characters
-/extensions
-/loras
-/models
-/presets
-/prompts
-/softprompts
-/training

.env.example

@@ -1,25 +0,0 @@
-# by default the Dockerfile specifies these versions: 3.5;5.0;6.0;6.1;7.0;7.5;8.0;8.6+PTX
-# however for me to work i had to specify the exact version for my card ( 2060 ) it was 7.5
-# https://developer.nvidia.com/cuda-gpus you can find the version for your card here
-TORCH_CUDA_ARCH_LIST=7.5
-
-# these commands worked for me with roughly 4.5GB of vram
-CLI_ARGS=--model llama-7b-4bit --wbits 4 --listen --auto-devices
-
-# the following examples have been tested with the files linked in docs/README_docker.md:
-# example running 13b with 4bit/128 groupsize        : CLI_ARGS=--model llama-13b-4bit-128g --wbits 4 --listen --groupsize 128 --pre_layer 25
-# example with loading api extension and public share: CLI_ARGS=--model llama-7b-4bit --wbits 4 --listen --auto-devices --no-stream --extensions api --share
-# example running 7b with 8bit groupsize             : CLI_ARGS=--model llama-7b --load-in-8bit --listen --auto-devices
-
-# the port the webui binds to on the host
-HOST_PORT=7860
-# the port the webui binds to inside the container
-CONTAINER_PORT=7860
-
-# the port the api binds to on the host
-HOST_API_PORT=5000
-# the port the api binds to inside the container
-CONTAINER_API_PORT=5000
-
-# the version used to install text-generation-webui from
-WEBUI_VERSION=HEAD

Dockerfile

@@ -1,61 +0,0 @@
-FROM nvidia/cuda:11.8.0-devel-ubuntu22.04 as builder
-
-RUN apt-get update && \
-    apt-get install --no-install-recommends -y git vim build-essential python3-dev python3-venv && \
-    rm -rf /var/lib/apt/lists/*
-
-RUN git clone https://github.com/oobabooga/GPTQ-for-LLaMa /build
-
-WORKDIR /build
-
-RUN python3 -m venv /build/venv
-RUN . /build/venv/bin/activate && \
-    pip3 install --upgrade pip setuptools && \
-    pip3 install torch torchvision torchaudio && \
-    pip3 install -r requirements.txt
-
-# https://developer.nvidia.com/cuda-gpus
-# for a rtx 2060: ARG TORCH_CUDA_ARCH_LIST="7.5"
-ARG TORCH_CUDA_ARCH_LIST="3.5;5.0;6.0;6.1;7.0;7.5;8.0;8.6+PTX"
-RUN . /build/venv/bin/activate && \
-    python3 setup_cuda.py bdist_wheel -d .
-
-FROM nvidia/cuda:11.8.0-runtime-ubuntu22.04
-
-LABEL maintainer="Your Name <your.email@example.com>"
-LABEL description="Docker image for GPTQ-for-LLaMa and Text Generation WebUI"
-
-RUN apt-get update && \
-    apt-get install --no-install-recommends -y git python3 python3-pip && \
-    rm -rf /var/lib/apt/lists/*
-
-RUN --mount=type=cache,target=/root/.cache/pip pip3 install virtualenv
-
-COPY . /app/
-
-WORKDIR /app
-
-ARG WEBUI_VERSION
-RUN test -n "${WEBUI_VERSION}" && git reset --hard ${WEBUI_VERSION} || echo "Using provided webui source"
-
-RUN virtualenv /app/venv
-RUN . /app/venv/bin/activate && \
-    pip3 install --upgrade pip setuptools && \
-    pip3 install torch torchvision torchaudio && \
-    pip3 install -r requirements.txt
-
-COPY --from=builder /build /app/repositories/GPTQ-for-LLaMa
-RUN . /app/venv/bin/activate && \
-    pip3 install /app/repositories/GPTQ-for-LLaMa/*.whl
-
-ENV CLI_ARGS=""
-
-RUN --mount=type=cache,target=/root/.cache/pip . /app/venv/bin/activate && cd extensions/api && pip3 install -r requirements.txt
-RUN --mount=type=cache,target=/root/.cache/pip . /app/venv/bin/activate && cd extensions/elevenlabs_tts && pip3 install -r requirements.txt
-RUN --mount=type=cache,target=/root/.cache/pip . /app/venv/bin/activate && cd extensions/google_translate && pip3 install -r requirements.txt
-RUN --mount=type=cache,target=/root/.cache/pip . /app/venv/bin/activate && cd extensions/silero_tts && pip3 install -r requirements.txt
-RUN --mount=type=cache,target=/root/.cache/pip . /app/venv/bin/activate && cd extensions/whisper_stt && pip3 install -r requirements.txt
-
-RUN cp /app/venv/lib/python3.10/site-packages/bitsandbytes/libbitsandbytes_cuda118.so /app/venv/lib/python3.10/site-packages/bitsandbytes/libbitsandbytes_cpu.so
-
-CMD . /app/venv/bin/activate && python3 server.py ${CLI_ARGS}

README.md

@@ -15,7 +15,6 @@ Its goal is to become the [AUTOMATIC1111/stable-diffusion-webui](https://github.
 * Dropdown menu for switching between models
 * Notebook mode that resembles OpenAI's playground
 * Chat mode for conversation and role playing
-* Instruct mode compatible with Alpaca and Open Assistant formats **\*NEW!\***
 * Nice HTML output for GPT-4chan
 * Markdown output for [GALACTICA](https://github.com/paperswithcode/galai), including LaTeX rendering
 * [Custom chat characters](https://github.com/oobabooga/text-generation-webui/wiki/Custom-chat-characters)
@@ -31,7 +30,7 @@ Its goal is to become the [AUTOMATIC1111/stable-diffusion-webui](https://github.
 * [LLaMA model, including 4-bit GPTQ](https://github.com/oobabooga/text-generation-webui/wiki/LLaMA-model)
 * [llama.cpp](https://github.com/oobabooga/text-generation-webui/wiki/llama.cpp-models) **\*NEW!\***
 * [RWKV model](https://github.com/oobabooga/text-generation-webui/wiki/RWKV-model)
-* [LoRA (loading and training)](https://github.com/oobabooga/text-generation-webui/wiki/Using-LoRAs)
+* [LoRa (loading and training)](https://github.com/oobabooga/text-generation-webui/wiki/Using-LoRAs)
 * Softprompts
 * [Extensions](https://github.com/oobabooga/text-generation-webui/wiki/Extensions)
 * [Google Colab](https://github.com/oobabooga/text-generation-webui/wiki/Running-on-Colab)
@@ -117,26 +116,8 @@ As an alternative to the recommended WSL method, you can install the web UI nati
 
 ### Alternative: Docker
 
-```
-cp .env.example .env
-docker-compose up --build
-```
+https://github.com/oobabooga/text-generation-webui/issues/174, https://github.com/oobabooga/text-generation-webui/issues/87
 
-Make sure to edit `.env.example` and set the appropriate CUDA version for your GPU.
-
-You need to have docker compose v2.17 or higher installed in your system. For installation instructions, see [Docker compose installation](https://github.com/oobabooga/text-generation-webui/wiki/Docker-compose-installation).
-
-Contributed by [@loeken](https://github.com/loeken) in [#633](https://github.com/oobabooga/text-generation-webui/pull/633)
-
-### Updating the requirements
-
-From time to time, the `requirements.txt` changes. To update, use this command:
-
-```
-conda activate textgen
-cd text-generation-webui
-pip install -r requirements.txt --upgrade
-```
 ## Downloading models
 
 Models should be placed inside the `models` folder.

api-example-stream.py

@@ -17,7 +17,6 @@ def random_hash():
     letters = string.ascii_lowercase + string.digits
     return ''.join(random.choice(letters) for i in range(9))
 
-
 async def run(context):
     server = "127.0.0.1"
     params = {
@@ -70,7 +69,6 @@ async def run(context):
 
 prompt = "What I would like to say is the following: "
 
-
 async def get_result():
     async for response in run(prompt):
         # Print intermediate steps

characters/instruction-following/Vicuna.yaml

@@ -1,3 +0,0 @@
-name: "### Assistant:"
-your_name: "### Human:"
-context: "Below is an instruction that describes a task. Write a response that appropriately completes the request."

convert-to-flexgen.py

@@ -17,7 +17,6 @@ parser = argparse.ArgumentParser(formatter_class=lambda prog: argparse.HelpForma
 parser.add_argument('MODEL', type=str, default=None, nargs='?', help="Path to the input model.")
 args = parser.parse_args()
 
-
 def disable_torch_init():
     """
     Disable the redundant torch default initialization to accelerate model creation.
@@ -32,14 +31,12 @@ def disable_torch_init():
     torch_layer_norm_init_backup = torch.nn.LayerNorm.reset_parameters
     setattr(torch.nn.LayerNorm, "reset_parameters", lambda self: None)
 
-
 def restore_torch_init():
     """Rollback the change made by disable_torch_init."""
     import torch
     setattr(torch.nn.Linear, "reset_parameters", torch_linear_init_backup)
     setattr(torch.nn.LayerNorm, "reset_parameters", torch_layer_norm_init_backup)
 
-
 if __name__ == '__main__':
     path = Path(args.MODEL)
     model_name = path.name

css/html_cai_style.css

@@ -64,15 +64,6 @@
     line-height: 1.428571429 !important;
 }
 
-.message-body li {
-    margin-top: 0.5em !important;
-    margin-bottom: 0.5em !important;
-}
-
-.message-body li > p {
-    display: inline !important;
-}
-
 .dark .message-body p em {
     color: rgb(138, 138, 138) !important;
 }

css/html_instruct_style.css

@@ -18,6 +18,10 @@
     line-height: 1.428571429;
 }
 
+.text p {
+    margin-top: 5px;
+}
+
 .username {
     display: none;
 }
@@ -30,15 +34,6 @@
     line-height: 1.428571429 !important;
 }
 
-.message-body li {
-    margin-top: 0.5em !important;
-    margin-bottom: 0.5em !important;
-}
-
-.message-body li > p {
-    display: inline !important;
-}
-
 .dark .message-body p em {
     color: rgb(138, 138, 138) !important;
 }
@@ -47,19 +42,15 @@
     color: rgb(110, 110, 110) !important;
 }
 
-.gradio-container .chat .assistant-message {
-  padding: 15px;
-  border-radius: 20px;
-  background-color: #0000000f;
-  margin-bottom: 17.5px;
+.assistant-message {
+  padding: 10px;
 }
 
-.gradio-container .chat .user-message {
-  padding: 15px;
-  border-radius: 20px;
-  margin-bottom: 17.5px !important;
+.user-message {
+  padding: 10px;
+  background-color: #f1f1f1;
 }
 
-.dark .chat .assistant-message {
-  background-color: #ffffff21;
+.dark .user-message {
+  background-color: #ffffff1a;
 }

css/main.css

@@ -41,7 +41,7 @@ ol li p, ul li p {
     display: inline-block;
 }
 
-#main, #parameters, #chat-settings, #interface-mode, #lora, #training-tab, #model-tab {
+#main, #parameters, #chat-settings, #interface-mode, #lora, #training-tab {
   border: 0;
 }
 

docker-compose.yml

@@ -1,32 +0,0 @@
-version: "3.3"
-services:
-  text-generation-webui:
-    build:
-      context: .
-      args:
-        # specify which cuda version your card supports: https://developer.nvidia.com/cuda-gpus
-        TORCH_CUDA_ARCH_LIST: ${TORCH_CUDA_ARCH_LIST}
-        GPTQ_VERSION: ${GPTQ_VERSION}
-        WEBUI_VERSION: ${WEBUI_VERSION}
-    env_file: .env
-    ports:
-      - "${HOST_PORT}:${CONTAINER_PORT}"
-      - "${HOST_API_PORT}:${CONTAINER_API_PORT}"
-    stdin_open: true
-    tty: true
-    volumes:
-      - ./characters:/app/characters
-      - ./extensions:/app/extensions
-      - ./loras:/app/loras
-      - ./models:/app/models
-      - ./presets:/app/presets
-      - ./prompts:/app/prompts
-      - ./softprompts:/app/softprompts
-      - ./training:/app/training
-    deploy:
-      resources:
-        reservations:
-          devices:
-            - driver: nvidia
-              device_ids: ['0']
-              capabilities: [gpu]

download-model.py

@@ -29,7 +29,6 @@ parser.add_argument('--clean', action='store_true', help='Does not resume the pr
 parser.add_argument('--check', action='store_true', help='Validates the checksums of model files.')
 args = parser.parse_args()
 
-
 def get_file(url, output_folder):
     filename = Path(url.rsplit('/', 1)[1])
     output_path = output_folder / filename
@@ -55,7 +54,6 @@ def get_file(url, output_folder):
                 t.update(len(data))
                 f.write(data)
 
-
 def sanitize_branch_name(branch_name):
     pattern = re.compile(r"^[a-zA-Z0-9._-]+$")
     if pattern.match(branch_name):
@@ -63,7 +61,6 @@ def sanitize_branch_name(branch_name):
     else:
         raise ValueError("Invalid branch name. Only alphanumeric characters, period, underscore and dash are allowed.")
 
-
 def select_model_from_default_options():
     models = {
         "OPT 6.7B": ("facebook", "opt-6.7b", "main"),
@@ -109,7 +106,6 @@ EleutherAI/pythia-1.4b-deduped
 
     return model, branch
 
-
 def get_download_links_from_huggingface(model, branch):
     base = "https://huggingface.co"
     page = f"/api/models/{model}/tree/{branch}?cursor="
@@ -176,11 +172,9 @@ def get_download_links_from_huggingface(model, branch):
 
     return links, sha256, is_lora
 
-
 def download_files(file_list, output_folder, num_threads=8):
     thread_map(lambda url: get_file(url, output_folder), file_list, max_workers=num_threads, disable=True)
 
-
 if __name__ == '__main__':
     model = args.MODEL
     branch = args.branch

extensions/api/script.py

@@ -9,7 +9,6 @@ params = {
     'port': 5000,
 }
 
-
 class Handler(BaseHTTPRequestHandler):
     def do_GET(self):
         if self.path == '/api/v1/model':
@@ -33,7 +32,7 @@ class Handler(BaseHTTPRequestHandler):
             self.end_headers()
 
             prompt = body['prompt']
-            prompt_lines = [k.strip() for k in prompt.split('\n')]
+            prompt_lines = [l.strip() for l in prompt.split('\n')]
 
             max_context = body.get('max_context_length', 2048)
 
@@ -96,6 +95,5 @@ def run_server():
         print(f'Starting KoboldAI compatible api at http://{server_addr[0]}:{server_addr[1]}/api')
     server.serve_forever()
 
-
 def setup():
     Thread(target=run_server, daemon=True).start()

extensions/character_bias/script.py

@@ -5,7 +5,6 @@ params = {
     "bias string": " *I am so happy*",
 }
 
-
 def input_modifier(string):
     """
     This function is applied to your text inputs before
@@ -14,7 +13,6 @@ def input_modifier(string):
 
     return string
 
-
 def output_modifier(string):
     """
     This function is applied to the model outputs.
@@ -22,7 +20,6 @@ def output_modifier(string):
 
     return string
 
-
 def bot_prefix_modifier(string):
     """
     This function is only applied in chat mode. It modifies
@@ -30,12 +27,11 @@ def bot_prefix_modifier(string):
     behavior.
     """
 
-    if params['activate']:
+    if params['activate'] == True:
         return f'{string} {params["bias string"].strip()} '
     else:
         return string
 
-
 def ui():
     # Gradio elements
     activate = gr.Checkbox(value=params['activate'], label='Activate character bias')

extensions/elevenlabs_tts/script.py

@@ -22,8 +22,6 @@ if not shared.args.no_stream:
     raise ValueError
     
 # Check if the API is valid and refresh the UI accordingly.
-
-
 def check_valid_api():
     
     global user, user_info, params
@@ -31,7 +29,7 @@ def check_valid_api():
     user = ElevenLabsUser(params['api_key'])
     user_info = user._get_subscription_data()
     print('checking api')
-    if not params['activate']:
+    if params['activate'] == False:
         return gr.update(value='Disconnected')
     elif user_info is None:
         print('Incorrect API Key')
@@ -41,8 +39,6 @@ def check_valid_api():
         return gr.update(value='Connected')
     
 # Once the API is verified, get the available voices and update the dropdown list
-
-
 def refresh_voices():
     
     global user, user_info
@@ -55,13 +51,11 @@ def refresh_voices():
     else:
         return
 
-
 def remove_surrounded_chars(string):
     # this expression matches to 'as few symbols as possible (0 upwards) between any asterisks' OR
     # 'as few symbols as possible (0 upwards) between an asterisk and the end of the string'
     return re.sub('\*[^\*]*?(\*|$)','',string)
 
-
 def input_modifier(string):
     """
     This function is applied to your text inputs before
@@ -70,7 +64,6 @@ def input_modifier(string):
 
     return string
 
-
 def output_modifier(string):
     """
     This function is applied to the model outputs.
@@ -78,9 +71,9 @@ def output_modifier(string):
 
     global params, wav_idx, user, user_info
     
-    if not params['activate']:
+    if params['activate'] == False:
         return string
-    elif user_info is None:
+    elif user_info == None:
         return string
 
     string = remove_surrounded_chars(string)
@@ -101,7 +94,6 @@ def output_modifier(string):
     wav_idx += 1
     return string
 
-
 def ui():
 
     # Gradio elements

extensions/google_translate/script.py

@@ -7,7 +7,6 @@ params = {
 
 language_codes = {'Afrikaans': 'af', 'Albanian': 'sq', 'Amharic': 'am', 'Arabic': 'ar', 'Armenian': 'hy', 'Azerbaijani': 'az', 'Basque': 'eu', 'Belarusian': 'be', 'Bengali': 'bn', 'Bosnian': 'bs', 'Bulgarian': 'bg', 'Catalan': 'ca', 'Cebuano': 'ceb', 'Chinese (Simplified)': 'zh-CN', 'Chinese (Traditional)': 'zh-TW', 'Corsican': 'co', 'Croatian': 'hr', 'Czech': 'cs', 'Danish': 'da', 'Dutch': 'nl', 'English': 'en', 'Esperanto': 'eo', 'Estonian': 'et', 'Finnish': 'fi', 'French': 'fr', 'Frisian': 'fy', 'Galician': 'gl', 'Georgian': 'ka', 'German': 'de', 'Greek': 'el', 'Gujarati': 'gu', 'Haitian Creole': 'ht', 'Hausa': 'ha', 'Hawaiian': 'haw', 'Hebrew': 'iw', 'Hindi': 'hi', 'Hmong': 'hmn', 'Hungarian': 'hu', 'Icelandic': 'is', 'Igbo': 'ig', 'Indonesian': 'id', 'Irish': 'ga', 'Italian': 'it', 'Japanese': 'ja', 'Javanese': 'jw', 'Kannada': 'kn', 'Kazakh': 'kk', 'Khmer': 'km', 'Korean': 'ko', 'Kurdish': 'ku', 'Kyrgyz': 'ky', 'Lao': 'lo', 'Latin': 'la', 'Latvian': 'lv', 'Lithuanian': 'lt', 'Luxembourgish': 'lb', 'Macedonian': 'mk', 'Malagasy': 'mg', 'Malay': 'ms', 'Malayalam': 'ml', 'Maltese': 'mt', 'Maori': 'mi', 'Marathi': 'mr', 'Mongolian': 'mn', 'Myanmar (Burmese)': 'my', 'Nepali': 'ne', 'Norwegian': 'no', 'Nyanja (Chichewa)': 'ny', 'Pashto': 'ps', 'Persian': 'fa', 'Polish': 'pl', 'Portuguese (Portugal, Brazil)': 'pt', 'Punjabi': 'pa', 'Romanian': 'ro', 'Russian': 'ru', 'Samoan': 'sm', 'Scots Gaelic': 'gd', 'Serbian': 'sr', 'Sesotho': 'st', 'Shona': 'sn', 'Sindhi': 'sd', 'Sinhala (Sinhalese)': 'si', 'Slovak': 'sk', 'Slovenian': 'sl', 'Somali': 'so', 'Spanish': 'es', 'Sundanese': 'su', 'Swahili': 'sw', 'Swedish': 'sv', 'Tagalog (Filipino)': 'tl', 'Tajik': 'tg', 'Tamil': 'ta', 'Telugu': 'te', 'Thai': 'th', 'Turkish': 'tr', 'Ukrainian': 'uk', 'Urdu': 'ur', 'Uzbek': 'uz', 'Vietnamese': 'vi', 'Welsh': 'cy', 'Xhosa': 'xh', 'Yiddish': 'yi', 'Yoruba': 'yo', 'Zulu': 'zu'}
 
-
 def input_modifier(string):
     """
     This function is applied to your text inputs before
@@ -16,7 +15,6 @@ def input_modifier(string):
 
     return GoogleTranslator(source=params['language string'], target='en').translate(string)
 
-
 def output_modifier(string):
     """
     This function is applied to the model outputs.
@@ -24,7 +22,6 @@ def output_modifier(string):
 
     return GoogleTranslator(source='en', target=params['language string']).translate(string)
 
-
 def bot_prefix_modifier(string):
     """
     This function is only applied in chat mode. It modifies
@@ -34,7 +31,6 @@ def bot_prefix_modifier(string):
 
     return string
 
-
 def ui():
     # Finding the language name from the language code to use as the default value
     language_name = list(language_codes.keys())[list(language_codes.values()).index(params['language string'])]

extensions/llama_prompts/script.py

@@ -4,14 +4,12 @@ import pandas as pd
 
 df = pd.read_csv("https://raw.githubusercontent.com/devbrones/llama-prompts/main/prompts/prompts.csv")
 
-
 def get_prompt_by_name(name):
     if name == 'None':
         return ''
     else:
         return df[df['Prompt name'] == name].iloc[0]['Prompt'].replace('\\n', '\n')
 
-
 def ui():
     if not shared.is_chat():
         choices = ['None'] + list(df['Prompt name'])

extensions/sd_api_pictures/script.py

@@ -30,15 +30,12 @@ streaming_state = shared.args.no_stream  # remember if chat streaming was enable
 picture_response = False # specifies if the next model response should appear as a picture
 pic_id = 0
 
-
 def remove_surrounded_chars(string):
     # this expression matches to 'as few symbols as possible (0 upwards) between any asterisks' OR
     # 'as few symbols as possible (0 upwards) between an asterisk and the end of the string'
     return re.sub('\*[^\*]*?(\*|$)','',string)
 
 # I don't even need input_hijack for this as visible text will be commited to history as the unmodified string
-
-
 def input_modifier(string):
     """
     This function is applied to your text inputs before
@@ -65,8 +62,6 @@ def input_modifier(string):
     return string
 
 # Get and save the Stable Diffusion-generated picture
-
-
 def get_SD_pictures(description):
 
     global params, pic_id
@@ -106,8 +101,6 @@ def get_SD_pictures(description):
 
 # TODO: how do I make the UI history ignore the resulting pictures (I don't want HTML to appear in history)
 # and replace it with 'text' for the purposes of logging?
-
-
 def output_modifier(string):
     """
     This function is applied to the model outputs.
@@ -137,7 +130,6 @@ def output_modifier(string):
     shared.args.no_stream = streaming_state
     return image + "\n" + text
 
-
 def bot_prefix_modifier(string):
     """
     This function is only applied in chat mode. It modifies
@@ -147,12 +139,10 @@ def bot_prefix_modifier(string):
 
     return string
 
-
 def force_pic():
     global picture_response
     picture_response = True
 
-
 def ui():
 
     # Gradio elements

extensions/send_pictures/script.py

@@ -17,13 +17,11 @@ input_hijack = {
 processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
 model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base", torch_dtype=torch.float32).to("cpu")
 
-
 def caption_image(raw_image):
     inputs = processor(raw_image.convert('RGB'), return_tensors="pt").to("cpu", torch.float32)
     out = model.generate(**inputs, max_new_tokens=100)
     return processor.decode(out[0], skip_special_tokens=True)
 
-
 def generate_chat_picture(picture, name1, name2):
     text = f'*{name1} sends {name2} a picture that contains the following: "{caption_image(picture)}"*'
     # lower the resolution of sent images for the chat, otherwise the log size gets out of control quickly with all the base64 values in visible history
@@ -34,7 +32,6 @@ def generate_chat_picture(picture, name1, name2):
     visible_text = f'<img src="data:image/jpeg;base64,{img_str}" alt="{text}">'
     return text, visible_text
 
-
 def ui():
     picture_select = gr.Image(label='Send a picture', type='pil')
 

modules/GPTQ_loader.py

@@ -1,4 +1,3 @@
-import inspect
 import re
 import sys
 from pathlib import Path
@@ -17,11 +16,9 @@ from quant import make_quant
 
 
 def _load_quant(model, checkpoint, wbits, groupsize=-1, faster_kernel=False, exclude_layers=['lm_head'], kernel_switch_threshold=128):
-
+    config = AutoConfig.from_pretrained(model)
     def noop(*args, **kwargs):
         pass
-
-    config = AutoConfig.from_pretrained(model)
     torch.nn.init.kaiming_uniform_ = noop 
     torch.nn.init.uniform_ = noop 
     torch.nn.init.normal_ = noop 
@@ -36,37 +33,21 @@ def _load_quant(model, checkpoint, wbits, groupsize=-1, faster_kernel=False, exc
     for name in exclude_layers:
         if name in layers:
             del layers[name]
-
-    gptq_args = inspect.getfullargspec(make_quant).args
-
-    make_quant_kwargs = {
-        'module': model,
-        'names': layers,
-        'bits': wbits,
-    }
-    if 'groupsize' in gptq_args:
-        make_quant_kwargs['groupsize'] = groupsize
-    if 'faster' in gptq_args:
-        make_quant_kwargs['faster'] = faster_kernel
-    if 'kernel_switch_threshold' in gptq_args:
-        make_quant_kwargs['kernel_switch_threshold'] = kernel_switch_threshold
-
-    make_quant(**make_quant_kwargs)
+    make_quant(model, layers, wbits, groupsize, faster=faster_kernel, kernel_switch_threshold=kernel_switch_threshold)
 
     del layers
     
     print('Loading model ...')
     if checkpoint.endswith('.safetensors'):
         from safetensors.torch import load_file as safe_load
-        model.load_state_dict(safe_load(checkpoint), strict=False)
+        model.load_state_dict(safe_load(checkpoint))
     else:
-        model.load_state_dict(torch.load(checkpoint), strict=False)
+        model.load_state_dict(torch.load(checkpoint))
     model.seqlen = 2048
     print('Done.')
 
     return model
 
-
 def load_quantized(model_name):
     if not shared.args.model_type:
         # Try to determine model type from model name

modules/LoRA.py

@@ -13,7 +13,6 @@ def reload_model():
     clear_torch_cache()
     shared.model, shared.tokenizer = load_model(shared.model_name)
 
-
 def add_lora_to_model(lora_name):
 
     # If a LoRA had been previously loaded, or if we want

modules/RWKV.py

@@ -54,7 +54,6 @@ class RWKVModel:
                 reply += token
                 yield reply
 
-
 class RWKVTokenizer:
     def __init__(self):
         pass

modules/api.py

@@ -28,7 +28,6 @@ def generate_reply_wrapper(string):
     for i in generate_reply(params[0], generate_params):
         yield i
 
-
 def create_apis():
     t1 = gr.Textbox(visible=False)
     t2 = gr.Textbox(visible=False)

modules/callbacks.py

@@ -30,7 +30,6 @@ class _SentinelTokenStoppingCriteria(transformers.StoppingCriteria):
                         return True
         return False
 
-
 class Stream(transformers.StoppingCriteria):
     def __init__(self, callback_func=None):
         self.callback_func = callback_func
@@ -40,7 +39,6 @@ class Stream(transformers.StoppingCriteria):
             self.callback_func(input_ids[0])
         return False
 
-
 class Iteratorize:
 
     """
@@ -98,7 +96,6 @@ class Iteratorize:
         self.stop_now = True
         clear_torch_cache()
 
-
 def clear_torch_cache():
     gc.collect()
     if not shared.args.cpu:

modules/chat.py

@@ -23,6 +23,8 @@ def generate_chat_prompt(user_input, max_new_tokens, name1, name2, context, chat
     end_of_turn = kwargs['end_of_turn'] if 'end_of_turn' in kwargs else ''
     impersonate = kwargs['impersonate'] if 'impersonate' in kwargs else False
     also_return_rows = kwargs['also_return_rows'] if 'also_return_rows' in kwargs else False
+
+    user_input = fix_newlines(user_input)
     rows = [f"{context.strip()}\n"]
 
     # Finding the maximum prompt size
@@ -49,8 +51,8 @@ def generate_chat_prompt(user_input, max_new_tokens, name1, name2, context, chat
         rows.append(f"{prefix1.strip() if not is_instruct else prefix1}")
         limit = 2
     else:
+
         # Adding the user message
-        user_input = fix_newlines(user_input)
         if len(user_input) > 0:
             rows.append(f"{prefix1}{user_input}{end_of_turn}\n")
 
@@ -67,7 +69,6 @@ def generate_chat_prompt(user_input, max_new_tokens, name1, name2, context, chat
     else:
         return prompt
 
-
 def extract_message_from_reply(reply, name1, name2, stop_at_newline):
     next_character_found = False
 
@@ -91,20 +92,12 @@ def extract_message_from_reply(reply, name1, name2, stop_at_newline):
                     if reply[-j:] == string[:j]:
                         reply = reply[:-j]
                         break
-                else:
-                    continue
-                break
 
     reply = fix_newlines(reply)
     return reply, next_character_found
 
-
 def chatbot_wrapper(text, generate_state, name1, name2, context, mode, end_of_turn, regenerate=False):
-    if mode == 'instruct':
-        stopping_strings = [f"\n{name1}", f"\n{name2}"]
-    else:
-        stopping_strings = [f"\n{name1}:", f"\n{name2}:"]
-
+    just_started = True
     eos_token = '\n' if generate_state['stop_at_newline'] else None
     name1_original = name1
     if 'pygmalion' in shared.model_name.lower():
@@ -114,7 +107,7 @@ def chatbot_wrapper(text, generate_state, name1, name2, context, mode, end_of_tu
     visible_text = None
     custom_generate_chat_prompt = None
     for extension, _ in extensions_module.iterator():
-        if hasattr(extension, 'input_hijack') and extension.input_hijack['state']:
+        if hasattr(extension, 'input_hijack') and extension.input_hijack['state'] == True:
             extension.input_hijack['state'] = False
             text, visible_text = extension.input_hijack['value']
         if custom_generate_chat_prompt is None and hasattr(extension, 'custom_generate_chat_prompt'):
@@ -136,10 +129,9 @@ def chatbot_wrapper(text, generate_state, name1, name2, context, mode, end_of_tu
 
     # Generate
     cumulative_reply = ''
-    just_started = True
     for i in range(generate_state['chat_generation_attempts']):
         reply = None
-        for reply in generate_reply(f"{prompt}{' ' if len(cumulative_reply) > 0 else ''}{cumulative_reply}", generate_state, eos_token=eos_token, stopping_strings=stopping_strings):
+        for reply in generate_reply(f"{prompt}{' ' if len(cumulative_reply) > 0 else ''}{cumulative_reply}", generate_state, eos_token=eos_token, stopping_strings=[f"\n{name1}:", f"\n{name2}:"]):
             reply = cumulative_reply + reply
 
             # Extracting the reply
@@ -168,14 +160,9 @@ def chatbot_wrapper(text, generate_state, name1, name2, context, mode, end_of_tu
 
     yield shared.history['visible']
 
-
 def impersonate_wrapper(text, generate_state, name1, name2, context, mode, end_of_turn):
-    if mode == 'instruct':
-        stopping_strings = [f"\n{name1}", f"\n{name2}"]
-    else:
-        stopping_strings = [f"\n{name1}:", f"\n{name2}:"]
-
     eos_token = '\n' if generate_state['stop_at_newline'] else None
+
     if 'pygmalion' in shared.model_name.lower():
         name1 = "You"
 
@@ -187,7 +174,7 @@ def impersonate_wrapper(text, generate_state, name1, name2, context, mode, end_o
     cumulative_reply = ''
     for i in range(generate_state['chat_generation_attempts']):
         reply = None
-        for reply in generate_reply(f"{prompt}{' ' if len(cumulative_reply) > 0 else ''}{cumulative_reply}", generate_state, eos_token=eos_token, stopping_strings=stopping_strings):
+        for reply in generate_reply(f"{prompt}{' ' if len(cumulative_reply) > 0 else ''}{cumulative_reply}", generate_state, eos_token=eos_token, stopping_strings=[f"\n{name1}:", f"\n{name2}:"]):
             reply = cumulative_reply + reply
             reply, next_character_found = extract_message_from_reply(reply, name1, name2, generate_state['stop_at_newline'])
             yield reply
@@ -199,12 +186,10 @@ def impersonate_wrapper(text, generate_state, name1, name2, context, mode, end_o
 
     yield reply
 
-
 def cai_chatbot_wrapper(text, generate_state, name1, name2, context, mode, end_of_turn):
-    for history in chatbot_wrapper(text, generate_state, name1, name2, context, mode, end_of_turn):
+    for history in chatbot_wrapper(text, generate_state, name1, name2, context, mode, end_of_turn, regenerate=False):
         yield chat_html_wrapper(history, name1, name2, mode)
 
-
 def regenerate_wrapper(text, generate_state, name1, name2, context, mode, end_of_turn):
     if (shared.character != 'None' and len(shared.history['visible']) == 1) or len(shared.history['internal']) == 0:
         yield chat_html_wrapper(shared.history['visible'], name1, name2, mode)
@@ -217,7 +202,6 @@ def regenerate_wrapper(text, generate_state, name1, name2, context, mode, end_of
             shared.history['visible'][-1] = [last_visible[0], history[-1][1]]
             yield chat_html_wrapper(shared.history['visible'], name1, name2, mode)
 
-
 def remove_last_message(name1, name2, mode):
     if len(shared.history['visible']) > 0 and shared.history['internal'][-1][0] != '<|BEGIN-VISIBLE-CHAT|>':
         last = shared.history['visible'].pop()
@@ -227,14 +211,12 @@ def remove_last_message(name1, name2, mode):
 
     return chat_html_wrapper(shared.history['visible'], name1, name2, mode), last[0]
 
-
 def send_last_reply_to_input():
     if len(shared.history['internal']) > 0:
         return shared.history['internal'][-1][1]
     else:
         return ''
 
-
 def replace_last_reply(text, name1, name2, mode):
     if len(shared.history['visible']) > 0:
         shared.history['visible'][-1][1] = text
@@ -242,11 +224,9 @@ def replace_last_reply(text, name1, name2, mode):
 
     return chat_html_wrapper(shared.history['visible'], name1, name2, mode)
 
-
 def clear_html():
     return chat_html_wrapper([], "", "")
 
-
 def clear_chat_log(name1, name2, greeting, mode):
     shared.history['visible'] = []
     shared.history['internal'] = []
@@ -257,11 +237,9 @@ def clear_chat_log(name1, name2, greeting, mode):
 
     return chat_html_wrapper(shared.history['visible'], name1, name2, mode)
 
-
 def redraw_html(name1, name2, mode):
     return chat_html_wrapper(shared.history['visible'], name1, name2, mode)
 
-
 def tokenize_dialogue(dialogue, name1, name2, mode):
     history = []
 
@@ -299,7 +277,6 @@ def tokenize_dialogue(dialogue, name1, name2, mode):
 
     return history
 
-
 def save_history(timestamp=True):
     if timestamp:
         fname = f"{shared.character}_{datetime.now().strftime('%Y%m%d-%H%M%S')}.json"
@@ -311,7 +288,6 @@ def save_history(timestamp=True):
         f.write(json.dumps({'data': shared.history['internal'], 'data_visible': shared.history['visible']}, indent=2))
     return Path(f'logs/{fname}')
 
-
 def load_history(file, name1, name2):
     file = file.decode('utf-8')
     try:
@@ -336,12 +312,10 @@ def load_history(file, name1, name2):
         shared.history['internal'] = tokenize_dialogue(file, name1, name2)
         shared.history['visible'] = copy.deepcopy(shared.history['internal'])
 
-
 def replace_character_names(text, name1, name2):
     text = text.replace('{{user}}', name1).replace('{{char}}', name2)
     return text.replace('<USER>', name1).replace('<BOT>', name2)
 
-
 def build_pygmalion_style_context(data):
     context = ""
     if 'char_persona' in data and data['char_persona'] != '':
@@ -351,7 +325,6 @@ def build_pygmalion_style_context(data):
     context = f"{context.strip()}\n<START>\n"
     return context
 
-
 def generate_pfp_cache(character):
     cache_folder = Path("cache")
     if not cache_folder.exists():
@@ -364,7 +337,6 @@ def generate_pfp_cache(character):
             return img
     return None
 
-
 def load_character(character, name1, name2, mode):
     shared.character = character
     shared.history['internal'] = []
@@ -421,11 +393,9 @@ def load_character(character, name1, name2, mode):
 
     return name1, name2, picture, greeting, context, end_of_turn, chat_html_wrapper(shared.history['visible'], name1, name2, mode, reset_cache=True)
 
-
 def load_default_history(name1, name2):
     load_character("None", name1, name2, "chat")
 
-
 def upload_character(json_file, img, tavern=False):
     json_file = json_file if type(json_file) == str else json_file.decode('utf-8')
     data = json.loads(json_file)
@@ -444,7 +414,6 @@ def upload_character(json_file, img, tavern=False):
     print(f'New character saved to "characters/{outfile_name}.json".')
     return outfile_name
 
-
 def upload_tavern_character(img, name1, name2):
     _img = Image.open(io.BytesIO(img))
     _img.getexif()
@@ -453,13 +422,12 @@ def upload_tavern_character(img, name1, name2):
     _json = {"char_name": _json['name'], "char_persona": _json['description'], "char_greeting": _json["first_mes"], "example_dialogue": _json['mes_example'], "world_scenario": _json['scenario']}
     return upload_character(json.dumps(_json), img, tavern=True)
 
-
 def upload_your_profile_picture(img, name1, name2, mode):
     cache_folder = Path("cache")
     if not cache_folder.exists():
         cache_folder.mkdir()
 
-    if img is None:
+    if img == None:
         if Path("cache/pfp_me.png").exists():
             Path("cache/pfp_me.png").unlink()
     else:

modules/extensions.py

@@ -9,7 +9,6 @@ state = {}
 available_extensions = []
 setup_called = set()
 
-
 def load_extensions():
     global state
     for i, name in enumerate(shared.args.extensions):
@@ -24,16 +23,12 @@ def load_extensions():
                 traceback.print_exc()
 
 # This iterator returns the extensions in the order specified in the command-line
-
-
 def iterator():
     for name in sorted(state, key=lambda x : state[x][1]):
         if state[name][0] == True:
             yield eval(f"extensions.{name}.script"), name
 
 # Extension functions that map string -> string
-
-
 def apply_extensions(text, typ):
     for extension, _ in iterator():
         if typ == "input" and hasattr(extension, "input_modifier"):
@@ -44,7 +39,6 @@ def apply_extensions(text, typ):
             text = extension.bot_prefix_modifier(text)
     return text
 
-
 def create_extensions_block():
     global setup_called
 

modules/html_generator.py

@@ -24,7 +24,6 @@ with open(Path(__file__).resolve().parent / '../css/html_cai_style.css', 'r') as
 with open(Path(__file__).resolve().parent / '../css/html_instruct_style.css', 'r') as f:
     instruct_css = f.read()
 
-
 def fix_newlines(string):
     string = string.replace('\n', '\n\n')
     string = re.sub(r"\n{3,}", "\n\n", string)
@@ -32,8 +31,6 @@ def fix_newlines(string):
     return string
 
 # This could probably be generalized and improved
-
-
 def convert_to_markdown(string):
     string = string.replace('\\begin{code}', '```')
     string = string.replace('\\end{code}', '```')
@@ -43,13 +40,11 @@ def convert_to_markdown(string):
     string = fix_newlines(string)
     return markdown.markdown(string, extensions=['fenced_code']) 
 
-
 def generate_basic_html(string):
     string = convert_to_markdown(string)
     string = f'<style>{readable_css}</style><div class="container">{string}</div>'
     return string
 
-
 def process_post(post, c):
     t = post.split('\n')
     number = t[0].split(' ')[1]
@@ -64,7 +59,6 @@ def process_post(post, c):
     src = f'<span class="name">Anonymous </span> <span class="number">No.{number}</span>\n{src}'
     return src
 
-
 def generate_4chan_html(f):
     posts = []
     post = ''
@@ -104,7 +98,6 @@ def generate_4chan_html(f):
 
     return output
 
-
 def make_thumbnail(image):
     image = image.resize((350, round(image.size[1]/image.size[0]*350)), Image.Resampling.LANCZOS)
     if image.size[1] > 470:
@@ -112,7 +105,6 @@ def make_thumbnail(image):
 
     return image
 
-
 def get_image_cache(path):
     cache_folder = Path("cache")
     if not cache_folder.exists():
@@ -127,7 +119,6 @@ def get_image_cache(path):
 
     return image_cache[path][1]
 
-
 def generate_instruct_html(history):
     output = f'<style>{instruct_css}</style><div class="chat" id="chat">'
     for i,_row in enumerate(history[::-1]):
@@ -160,7 +151,6 @@ def generate_instruct_html(history):
 
     return output
 
-
 def generate_cai_chat_html(history, name1, name2, reset_cache=False):
     output = f'<style>{cai_css}</style><div class="chat" id="chat">'
 
@@ -210,11 +200,9 @@ def generate_cai_chat_html(history, name1, name2, reset_cache=False):
     output += "</div>"
     return output
 
-
 def generate_chat_html(history, name1, name2):
     return generate_cai_chat_html(history, name1, name2)
 
-
 def chat_html_wrapper(history, name1, name2, mode, reset_cache=False):
     if mode == "cai-chat":
         return generate_cai_chat_html(history, name1, name2, reset_cache)

modules/llamacpp_model_alternative.py

@@ -1,63 +0,0 @@
-'''
-Based on
-https://github.com/abetlen/llama-cpp-python
-
-Documentation:
-https://abetlen.github.io/llama-cpp-python/
-'''
-
-from llama_cpp import Llama
-
-from modules import shared
-from modules.callbacks import Iteratorize
-
-
-class LlamaCppModel:
-    def __init__(self):
-        self.initialized = False
-
-    @classmethod
-    def from_pretrained(self, path):
-        result = self()
-
-        params = {
-            'model_path': str(path),
-            'n_ctx': 2048,
-            'seed': 0,
-            'n_threads': shared.args.threads or None
-        }
-        self.model = Llama(**params)
-
-        # This is ugly, but the model and the tokenizer are the same object in this library.
-        return result, result
-
-    def encode(self, string):
-        if type(string) is str:
-            string = string.encode()
-        return self.model.tokenize(string)
-
-    def generate(self, context="", token_count=20, temperature=1, top_p=1, top_k=50, repetition_penalty=1, callback=None):
-        if type(context) is str:
-            context = context.encode()
-        tokens = self.model.tokenize(context)
-
-        output = b""
-        count = 0
-        for token in self.model.generate(tokens, top_k=top_k, top_p=top_p, temp=temperature, repeat_penalty=repetition_penalty):
-            text = self.model.detokenize([token])
-            output += text
-            if callback:
-                callback(text.decode())
-
-            count += 1
-            if count >= token_count or (token == self.model.token_eos()):
-                break
-
-        return output.decode()
-
-    def generate_with_streaming(self, **kwargs):
-        with Iteratorize(self.generate, kwargs, callback=None) as generator:
-            reply = ''
-            for token in generator:
-                reply += token
-                yield reply

modules/models.py

@@ -10,7 +10,7 @@ import torch
 import transformers
 from accelerate import infer_auto_device_map, init_empty_weights
 from transformers import (AutoConfig, AutoModelForCausalLM, AutoTokenizer,
-                          BitsAndBytesConfig, LlamaTokenizer)
+                          BitsAndBytesConfig)
 
 import modules.shared as shared
 
@@ -103,7 +103,7 @@ def load_model(model_name):
 
     # llamacpp model
     elif shared.is_llamacpp:
-        from modules.llamacpp_model_alternative import LlamaCppModel
+        from modules.llamacpp_model import LlamaCppModel
 
         model_file = list(Path(f'{shared.args.model_dir}/{model_name}').glob('ggml*.bin'))[0]
         print(f"llama.cpp weights detected: {model_file}\n")
@@ -172,8 +172,6 @@ def load_model(model_name):
     # Loading the tokenizer
     if any((k in shared.model_name.lower() for k in ['gpt4chan', 'gpt-4chan'])) and Path(f"{shared.args.model_dir}/gpt-j-6B/").exists():
         tokenizer = AutoTokenizer.from_pretrained(Path(f"{shared.args.model_dir}/gpt-j-6B/"))
-    elif type(model) is transformers.LlamaForCausalLM:
-        tokenizer = LlamaTokenizer.from_pretrained(Path(f"{shared.args.model_dir}/{shared.model_name}/"), clean_up_tokenization_spaces=True)
     else:
         tokenizer = AutoTokenizer.from_pretrained(Path(f"{shared.args.model_dir}/{shared.model_name}/"))
     tokenizer.truncation_side = 'left'
@@ -181,7 +179,6 @@ def load_model(model_name):
     print(f"Loaded the model in {(time.time()-t0):.2f} seconds.")
     return model, tokenizer
 
-
 def load_soft_prompt(name):
     if name == 'None':
         shared.soft_prompt = False

modules/shared.py

@@ -61,7 +61,6 @@ settings = {
     }
 }
 
-
 def str2bool(v):
     if isinstance(v, bool):
         return v
@@ -72,7 +71,6 @@ def str2bool(v):
     else:
         raise argparse.ArgumentTypeError('Boolean value expected.')
 
-
 parser = argparse.ArgumentParser(formatter_class=lambda prog: argparse.HelpFormatter(prog,max_help_position=54))
 
 # Basic settings
@@ -147,6 +145,5 @@ if args.cai_chat:
     print("Warning: --cai-chat is deprecated. Use --chat instead.")
     args.chat = True
 
-
 def is_chat():
     return args.chat

modules/text_generation.py

@@ -21,7 +21,6 @@ def get_max_prompt_length(tokens):
         max_length -= shared.soft_prompt_tensor.shape[1]
     return max_length
 
-
 def encode(prompt, tokens_to_generate=0, add_special_tokens=True):
     if any((shared.is_RWKV, shared.is_llamacpp)):
         input_ids = shared.tokenizer.encode(str(prompt))
@@ -29,10 +28,6 @@ def encode(prompt, tokens_to_generate=0, add_special_tokens=True):
         return input_ids
     else:
         input_ids = shared.tokenizer.encode(str(prompt), return_tensors='pt', truncation=True, max_length=get_max_prompt_length(tokens_to_generate), add_special_tokens=add_special_tokens)
-
-        if type(shared.tokenizer) is transformers.LlamaTokenizer and input_ids[0][0] == 29871:
-            input_ids = input_ids[:, 1:]
-
         if shared.args.cpu:
             return input_ids
         elif shared.args.flexgen:
@@ -45,7 +40,6 @@ def encode(prompt, tokens_to_generate=0, add_special_tokens=True):
         else:
             return input_ids.cuda()
 
-
 def decode(output_ids):
     # Open Assistant relies on special tokens like <|endoftext|>
     if re.match('.*(oasst|galactica)-*', shared.model_name.lower()):
@@ -55,7 +49,6 @@ def decode(output_ids):
         reply = reply.replace(r'<|endoftext|>', '')
         return reply
 
-
 def generate_softprompt_input_tensors(input_ids):
     inputs_embeds = shared.model.transformer.wte(input_ids)
     inputs_embeds = torch.cat((shared.soft_prompt_tensor, inputs_embeds), dim=1)
@@ -64,8 +57,6 @@ def generate_softprompt_input_tensors(input_ids):
     return inputs_embeds, filler_input_ids
 
 # Removes empty replies from gpt4chan outputs
-
-
 def fix_gpt4chan(s):
     for i in range(10):
         s = re.sub("--- [0-9]*\n>>[0-9]*\n---", "---", s)
@@ -74,8 +65,6 @@ def fix_gpt4chan(s):
     return s
 
 # Fix the LaTeX equations in galactica
-
-
 def fix_galactica(s):
     s = s.replace(r'\[', r'$')
     s = s.replace(r'\]', r'$')
@@ -86,7 +75,6 @@ def fix_galactica(s):
     s = re.sub(r"\n{3,}", "\n\n", s)
     return s
 
-
 def formatted_outputs(reply, model_name):
     if not shared.is_chat():
         if 'galactica' in model_name.lower():
@@ -100,24 +88,20 @@ def formatted_outputs(reply, model_name):
     else:
         return reply
 
-
 def clear_torch_cache():
     gc.collect()
     if not shared.args.cpu:
         torch.cuda.empty_cache()
 
-
 def set_manual_seed(seed):
     if seed != -1:
         torch.manual_seed(seed)
         if torch.cuda.is_available():
             torch.cuda.manual_seed_all(seed)
 
-
 def stop_everything_event():
     shared.stop_everything = True
 
-
 def generate_reply(question, generate_state, eos_token=None, stopping_strings=[]):
     clear_torch_cache()
     set_manual_seed(generate_state['seed'])

modules/training.py

@@ -19,10 +19,8 @@ CURRENT_STEPS = 0
 MAX_STEPS = 0
 CURRENT_GRADIENT_ACCUM = 1
 
-
 def get_dataset(path: str, ext: str):
-    return ['None'] + sorted(set([k.stem for k in Path(path).glob(f'*.{ext}') if k.stem != 'put-trainer-datasets-here']), key=str.lower)
-
+    return ['None'] + sorted(set((k.stem for k in Path(path).glob(f'*.{ext}'))), key=str.lower)
 
 def create_train_interface():
     with gr.Tab('Train LoRA', elem_id='lora-train-tab'):
@@ -47,34 +45,28 @@ def create_train_interface():
             with gr.Row():
                 dataset = gr.Dropdown(choices=get_dataset('training/datasets', 'json'), value='None', label='Dataset', info='The dataset file to use for training.')
                 ui.create_refresh_button(dataset, lambda : None, lambda : {'choices': get_dataset('training/datasets', 'json')}, 'refresh-button')
-                eval_dataset = gr.Dropdown(choices=get_dataset('training/datasets', 'json'), value='None', label='Evaluation Dataset', info='The (optional) dataset file used to evaluate the model after training.')
+                eval_dataset = gr.Dropdown(choices=get_dataset('training/datasets', 'json'), value='None', label='Evaluation Dataset', info='The dataset file used to evaluate the model after training.')
                 ui.create_refresh_button(eval_dataset, lambda : None, lambda : {'choices': get_dataset('training/datasets', 'json')}, 'refresh-button')
                 format = gr.Dropdown(choices=get_dataset('training/formats', 'json'), value='None', label='Data Format', info='The format file used to decide how to format the dataset input.')
                 ui.create_refresh_button(format, lambda : None, lambda : {'choices': get_dataset('training/formats', 'json')}, 'refresh-button')
-
         with gr.Tab(label="Raw Text File"):
             with gr.Row():
                 raw_text_file = gr.Dropdown(choices=get_dataset('training/datasets', 'txt'), value='None', label='Text File', info='The raw text file to use for training.')
                 ui.create_refresh_button(raw_text_file, lambda : None, lambda : {'choices': get_dataset('training/datasets', 'txt')}, 'refresh-button')
-            with gr.Row():
-                overlap_len = gr.Slider(label='Overlap Length', minimum=0, maximum=512, value=128, step=16, info='Overlap length - ie how many tokens from the prior chunk of text to include into the next chunk. (The chunks themselves will be of a size determined by Cutoff Length below). Setting overlap to exactly half the cutoff length may be ideal.')
-                newline_favor_len = gr.Slider(label='Prefer Newline Cut Length', minimum=0, maximum=512, value=128, step=16, info='Length (in characters, not tokens) of the maximum distance to shift an overlap cut by to ensure chunks cut at newlines. If too low, cuts may occur in the middle of lines.')
+                overlap_len = gr.Slider(label='Overlap Length', minimum=0,maximum=512, value=128, step=16, info='Overlap length - ie how many tokens from the prior chunk of text to include into the next chunk. (The chunks themselves will be of a size determined by Cutoff Length above). Setting overlap to exactly half the cutoff length may be ideal.')
 
         with gr.Row():
             start_button = gr.Button("Start LoRA Training")
             stop_button = gr.Button("Interrupt")
 
         output = gr.Markdown(value="Ready")
-        start_button.click(do_train, [lora_name, micro_batch_size, batch_size, epochs, learning_rate, lora_rank, lora_alpha, lora_dropout,
-                                      cutoff_len, dataset, eval_dataset, format, raw_text_file, overlap_len, newline_favor_len], [output])
+        start_button.click(do_train, [lora_name, micro_batch_size, batch_size, epochs, learning_rate, lora_rank, lora_alpha, lora_dropout, cutoff_len, dataset, eval_dataset, format, raw_text_file, overlap_len], [output])
         stop_button.click(do_interrupt, [], [], cancels=[], queue=False)
 
-
 def do_interrupt():
     global WANT_INTERRUPT
     WANT_INTERRUPT = True
 
-
 class Callbacks(transformers.TrainerCallback):
     def on_step_begin(self, args: transformers.TrainingArguments, state: transformers.TrainerState, control: transformers.TrainerControl, **kwargs):
         global CURRENT_STEPS, MAX_STEPS
@@ -83,7 +75,6 @@ class Callbacks(transformers.TrainerCallback):
         if WANT_INTERRUPT:
             control.should_epoch_stop = True
             control.should_training_stop = True
-
     def on_substep_end(self, args: transformers.TrainingArguments, state: transformers.TrainerState, control: transformers.TrainerControl, **kwargs):
         global CURRENT_STEPS
         CURRENT_STEPS += 1
@@ -91,7 +82,6 @@ class Callbacks(transformers.TrainerCallback):
             control.should_epoch_stop = True
             control.should_training_stop = True
 
-
 def clean_path(base_path: str, path: str):
     """"Strips unusual symbols and forcibly builds a path as relative to the intended directory."""
     # TODO: Probably could do with a security audit to guarantee there's no ways this can be bypassed to target an unwanted path.
@@ -101,9 +91,8 @@ def clean_path(base_path: str, path: str):
         return path
     return f'{Path(base_path).absolute()}/{path}'
 
-
-def do_train(lora_name: str, micro_batch_size: int, batch_size: int, epochs: int, learning_rate: str, lora_rank: int, lora_alpha: int, lora_dropout: float,
-             cutoff_len: int, dataset: str, eval_dataset: str, format: str, raw_text_file: str, overlap_len: int, newline_favor_len: int):
+def do_train(lora_name: str, micro_batch_size: int, batch_size: int, epochs: int, learning_rate: str, lora_rank: int,
+             lora_alpha: int, lora_dropout: float, cutoff_len: int, dataset: str, eval_dataset: str, format: str, raw_text_file: str, overlap_len: int):
     global WANT_INTERRUPT, CURRENT_STEPS, MAX_STEPS, CURRENT_GRADIENT_ACCUM
     WANT_INTERRUPT = False
     CURRENT_STEPS = 0
@@ -114,25 +103,6 @@ def do_train(lora_name: str, micro_batch_size: int, batch_size: int, epochs: int
     lora_name = f"{shared.args.lora_dir}/{clean_path(None, lora_name)}"
     actual_lr = float(learning_rate)
 
-    model_type = type(shared.model).__name__
-    if model_type != "LlamaForCausalLM":
-        if model_type == "PeftModelForCausalLM":
-            yield "You are trying to train a LoRA while you already have another LoRA loaded. This will work, but may have unexpected effects. *(Will continue anyway in 5 seconds, press `Interrupt` to stop.)*"
-            print("Warning: Training LoRA over top of another LoRA. May have unexpected effects.")
-        else:
-            yield "LoRA training has only currently been validated for LLaMA models. Unexpected errors may follow. *(Will continue anyway in 5 seconds, press `Interrupt` to stop.)*"
-            print(f"Warning: LoRA training has only currently been validated for LLaMA models. (Found model type: {model_type})")
-        time.sleep(5)
-
-    if shared.args.wbits > 0 or shared.args.gptq_bits > 0:
-        yield "LoRA training does not yet support 4bit. Please use `--load-in-8bit` for now."
-        return
-
-    elif not shared.args.load_in_8bit:
-        yield "It is highly recommended you use `--load-in-8bit` for LoRA training. *(Will continue anyway in 2 seconds, press `Interrupt` to stop.)*"
-        print("Warning: It is highly recommended you use `--load-in-8bit` for LoRA training.")
-        time.sleep(2)  # Give it a moment for the message to show in UI before continuing
-
     if cutoff_len <= 0 or micro_batch_size <= 0 or batch_size <= 0 or actual_lr <= 0 or lora_rank <= 0 or lora_alpha <= 0:
         yield "Cannot input zeroes."
         return
@@ -156,20 +126,15 @@ def do_train(lora_name: str, micro_batch_size: int, batch_size: int, epochs: int
             raw_text = file.read()
         tokens = shared.tokenizer.encode(raw_text)
         del raw_text # Note: could be a gig for a large dataset, so delete redundant data as we go to be safe on RAM
-
         tokens = list(split_chunks(tokens, cutoff_len - overlap_len))
         for i in range(1, len(tokens)):
             tokens[i] = tokens[i - 1][-overlap_len:] + tokens[i]
         text_chunks = [shared.tokenizer.decode(x) for x in tokens]
         del tokens
-
-        if newline_favor_len > 0:
-            text_chunks = [cut_chunk_for_newline(x, newline_favor_len) for x in text_chunks]
-
-        train_data = Dataset.from_list([tokenize(x) for x in text_chunks])
-        del text_chunks
-        train_data = train_data.shuffle()
+        data = Dataset.from_list([tokenize(x) for x in text_chunks])
+        train_data = data.shuffle()
         eval_data = None
+        del text_chunks
 
     else:
         if dataset in ['None', '']:
@@ -267,37 +232,33 @@ def do_train(lora_name: str, micro_batch_size: int, batch_size: int, epochs: int
     # TODO: save/load checkpoints to resume from?
     print("Starting training...")
     yield "Starting..."
-    if WANT_INTERRUPT:
-        yield "Interrupted before start."
-        return
 
-    def threaded_run():
+    def threadedRun():
         trainer.train()
 
-    thread = threading.Thread(target=threaded_run)
+    thread = threading.Thread(target=threadedRun)
     thread.start()
-    last_step = 0
-    start_time = time.perf_counter()
+    lastStep = 0
+    startTime = time.perf_counter()
 
     while thread.is_alive():
         time.sleep(0.5)
         if WANT_INTERRUPT:
             yield "Interrupting, please wait... *(Run will stop after the current training step completes.)*"
-
-        elif CURRENT_STEPS != last_step:
-            last_step = CURRENT_STEPS
-            time_elapsed = time.perf_counter() - start_time
-            if time_elapsed <= 0:
-                timer_info = ""
-                total_time_estimate = 999
+        elif CURRENT_STEPS != lastStep:
+            lastStep = CURRENT_STEPS
+            timeElapsed = time.perf_counter() - startTime
+            if timeElapsed <= 0:
+                timerInfo = ""
+                totalTimeEstimate = 999
             else:
-                its = CURRENT_STEPS / time_elapsed
+                its = CURRENT_STEPS / timeElapsed
                 if its > 1:
-                    timer_info = f"`{its:.2f}` it/s"
+                    timerInfo = f"`{its:.2f}` it/s"
                 else:
-                    timer_info = f"`{1.0/its:.2f}` s/it"
-                total_time_estimate = (1.0 / its) * (MAX_STEPS)
-            yield f"Running... **{CURRENT_STEPS}** / **{MAX_STEPS}** ... {timer_info}, {format_time(time_elapsed)} / {format_time(total_time_estimate)} ... {format_time(total_time_estimate - time_elapsed)} remaining"
+                    timerInfo = f"`{1.0/its:.2f}` s/it"
+                totalTimeEstimate = (1.0/its) * (MAX_STEPS)
+            yield f"Running... **{CURRENT_STEPS}** / **{MAX_STEPS}** ... {timerInfo}, `{timeElapsed:.0f}`/`{totalTimeEstimate:.0f}` seconds"
 
     print("Training complete, saving...")
     lora_model.save_pretrained(lora_name)
@@ -309,31 +270,6 @@ def do_train(lora_name: str, micro_batch_size: int, batch_size: int, epochs: int
         print("Training complete!")
         yield f"Done! LoRA saved to `{lora_name}`"
 
-
 def split_chunks(arr, step):
     for i in range(0, len(arr), step):
         yield arr[i:i + step]
-
-
-def cut_chunk_for_newline(chunk: str, max_length: int):
-    if '\n' not in chunk:
-        return chunk
-    first_newline = chunk.index('\n')
-    if first_newline < max_length:
-        chunk = chunk[first_newline + 1:]
-    if '\n' not in chunk:
-        return chunk
-    last_newline = chunk.rindex('\n')
-    if len(chunk) - last_newline < max_length:
-        chunk = chunk[:last_newline]
-    return chunk
-
-
-def format_time(seconds: float):
-    if seconds < 120:
-        return f"`{seconds:.0f}` seconds"
-    minutes = seconds / 60
-    if minutes < 120:
-        return f"`{minutes:.0f}` minutes"
-    hours = minutes / 60
-    return f"`{hours:.0f}` hours"

modules/ui.py

@@ -13,7 +13,6 @@ with open(Path(__file__).resolve().parent / '../css/main.js', 'r') as f:
 with open(Path(__file__).resolve().parent / '../css/chat.js', 'r') as f:
     chat_js = f.read()
 
-
 class ToolButton(gr.Button, gr.components.FormComponent):
     """Small button with single emoji as text, fits inside gradio forms"""
 
@@ -23,7 +22,6 @@ class ToolButton(gr.Button, gr.components.FormComponent):
     def get_block_name(self):
         return "button"
 
-
 def create_refresh_button(refresh_component, refresh_method, refreshed_args, elem_id):
     def refresh():
         refresh_method()

requirements.txt

@@ -3,13 +3,14 @@ bitsandbytes==0.37.2
 datasets
 flexgen==0.1.7
 gradio==3.24.1
+llamacpp==0.1.11
 markdown
 numpy
 peft==0.2.0
 requests
-rwkv==0.7.3
+rwkv==0.7.2
 safetensors==0.3.0
 sentencepiece
 pyyaml
 tqdm
-git+https://github.com/huggingface/transformers
+git+https://github.com/huggingface/transformers@9eae4aa57650c1dbe1becd4e0979f6ad1e572ac0

server.py

@@ -34,18 +34,15 @@ if settings_file is not None:
     for item in new_settings:
         shared.settings[item] = new_settings[item]
 
-
 def get_available_models():
     if shared.args.flexgen:
         return sorted([re.sub('-np$', '', item.name) for item in list(Path(f'{shared.args.model_dir}/').glob('*')) if item.name.endswith('-np')], key=str.lower)
     else:
         return sorted([re.sub('.pth$', '', item.name) for item in list(Path(f'{shared.args.model_dir}/').glob('*')) if not item.name.endswith(('.txt', '-np', '.pt', '.json'))], key=str.lower)
 
-
 def get_available_presets():
     return sorted(set((k.stem for k in Path('presets').glob('*.txt'))), key=str.lower)
 
-
 def get_available_prompts():
     prompts = []
     prompts += sorted(set((k.stem for k in Path('prompts').glob('[0-9]*.txt'))), key=str.lower, reverse=True)
@@ -53,37 +50,27 @@ def get_available_prompts():
     prompts += ['None']
     return prompts
 
-
 def get_available_characters():
     paths = (x for x in Path('characters').iterdir() if x.suffix in ('.json', '.yaml', '.yml'))
     return ['None'] + sorted(set((k.stem for k in paths if k.stem != "instruction-following")), key=str.lower)
 
-
 def get_available_instruction_templates():
-    path = "characters/instruction-following"
-    paths = []
-    if os.path.exists(path):
-        paths = (x for x in Path(path).iterdir() if x.suffix in ('.json', '.yaml', '.yml'))
+    paths = (x for x in Path('characters/instruction-following').iterdir() if x.suffix in ('.json', '.yaml', '.yml'))
     return ['None'] + sorted(set((k.stem for k in paths)), key=str.lower)
 
-
 def get_available_extensions():
     return sorted(set(map(lambda x : x.parts[1], Path('extensions').glob('*/script.py'))), key=str.lower)
 
-
 def get_available_softprompts():
     return ['None'] + sorted(set((k.stem for k in Path('softprompts').glob('*.zip'))), key=str.lower)
 
-
 def get_available_loras():
     return ['None'] + sorted([item.name for item in list(Path(shared.args.lora_dir).glob('*')) if not item.name.endswith(('.txt', '-np', '.pt', '.json'))], key=str.lower)
 
-
 def unload_model():
     shared.model = shared.tokenizer = None
     clear_torch_cache()
 
-
 def load_model_wrapper(selected_model):
     if selected_model != shared.model_name:
         shared.model_name = selected_model
@@ -94,12 +81,10 @@ def load_model_wrapper(selected_model):
 
     return selected_model
 
-
 def load_lora_wrapper(selected_lora):
     add_lora_to_model(selected_lora)
     return selected_lora
 
-
 def load_preset_values(preset_menu, state, return_dict=False):
     generate_params = {
         'do_sample': True,
@@ -130,7 +115,6 @@ def load_preset_values(preset_menu, state, return_dict=False):
         state.update(generate_params)
         return state, *[generate_params[k] for k in ['do_sample', 'temperature', 'top_p', 'typical_p', 'repetition_penalty', 'encoder_repetition_penalty', 'top_k', 'min_length', 'no_repeat_ngram_size', 'num_beams', 'penalty_alpha', 'length_penalty', 'early_stopping']]
 
-
 def upload_soft_prompt(file):
     with zipfile.ZipFile(io.BytesIO(file)) as zf:
         zf.extract('meta.json')
@@ -143,6 +127,16 @@ def upload_soft_prompt(file):
 
     return name
 
+def create_model_and_preset_menus():
+    with gr.Row():
+        with gr.Column():
+            with gr.Row():
+                shared.gradio['model_menu'] = gr.Dropdown(choices=available_models, value=shared.model_name, label='Model')
+                ui.create_refresh_button(shared.gradio['model_menu'], lambda : None, lambda : {'choices': get_available_models()}, 'refresh-button')
+        with gr.Column():
+            with gr.Row():
+                shared.gradio['preset_menu'] = gr.Dropdown(choices=available_presets, value=default_preset if not shared.args.flexgen else 'Naive', label='Generation parameters preset')
+                ui.create_refresh_button(shared.gradio['preset_menu'], lambda : None, lambda : {'choices': get_available_presets()}, 'refresh-button')
 
 def save_prompt(text):
     fname = f"{datetime.now().strftime('%Y-%m-%d-%H%M%S')}.txt"
@@ -150,7 +144,6 @@ def save_prompt(text):
         f.write(text)
     return f"Saved to prompts/{fname}"
 
-
 def load_prompt(fname):
     if fname in ['None', '']:
         return ''
@@ -161,7 +154,6 @@ def load_prompt(fname):
                 text = text[:-1]
             return text
 
-
 def create_prompt_menus():
     with gr.Row():
         with gr.Column():
@@ -177,22 +169,6 @@ def create_prompt_menus():
     shared.gradio['prompt_menu'].change(load_prompt, [shared.gradio['prompt_menu']], [shared.gradio['textbox']], show_progress=False)
     shared.gradio['save_prompt'].click(save_prompt, [shared.gradio['textbox']], [shared.gradio['status']], show_progress=False)
 
-
-def create_model_menus():
-    with gr.Row():
-        with gr.Column():
-            with gr.Row():
-                shared.gradio['model_menu'] = gr.Dropdown(choices=available_models, value=shared.model_name, label='Model')
-                ui.create_refresh_button(shared.gradio['model_menu'], lambda: None, lambda: {'choices': get_available_models()}, 'refresh-button')
-        with gr.Column():
-            with gr.Row():
-                shared.gradio['lora_menu'] = gr.Dropdown(choices=available_loras, value=shared.lora_name, label='LoRA')
-                ui.create_refresh_button(shared.gradio['lora_menu'], lambda: None, lambda: {'choices': get_available_loras()}, 'refresh-button')
-
-    shared.gradio['model_menu'].change(load_model_wrapper, shared.gradio['model_menu'], shared.gradio['model_menu'], show_progress=True)
-    shared.gradio['lora_menu'].change(load_lora_wrapper, shared.gradio['lora_menu'], shared.gradio['lora_menu'], show_progress=True)
-
-
 def create_settings_menus(default_preset):
     generate_params = load_preset_values(default_preset if not shared.args.flexgen else 'Naive', {}, return_dict=True)
     for k in ['max_new_tokens', 'seed', 'stop_at_newline', 'chat_prompt_size', 'chat_generation_attempts']:
@@ -201,9 +177,7 @@ def create_settings_menus(default_preset):
 
     with gr.Row():
         with gr.Column():
-            with gr.Row():
-                shared.gradio['preset_menu'] = gr.Dropdown(choices=available_presets, value=default_preset if not shared.args.flexgen else 'Naive', label='Generation parameters preset')
-                ui.create_refresh_button(shared.gradio['preset_menu'], lambda: None, lambda: {'choices': get_available_presets()}, 'refresh-button')
+            create_model_and_preset_menus()
         with gr.Column():
             shared.gradio['seed'] = gr.Number(value=shared.settings['seed'], label='Seed (-1 for random)')
 
@@ -227,6 +201,7 @@ def create_settings_menus(default_preset):
             with gr.Box():
                 gr.Markdown('Contrastive search')
                 shared.gradio['penalty_alpha'] = gr.Slider(0, 5, value=generate_params['penalty_alpha'], label='penalty_alpha')
+
             with gr.Box():
                 gr.Markdown('Beam search (uses a lot of VRAM)')
                 with gr.Row():
@@ -236,6 +211,10 @@ def create_settings_menus(default_preset):
                         shared.gradio['length_penalty'] = gr.Slider(-5, 5, value=generate_params['length_penalty'], label='length_penalty')
                 shared.gradio['early_stopping'] = gr.Checkbox(value=generate_params['early_stopping'], label='early_stopping')
 
+    with gr.Row():
+        shared.gradio['lora_menu'] = gr.Dropdown(choices=available_loras, value=shared.lora_name, label='LoRA')
+        ui.create_refresh_button(shared.gradio['lora_menu'], lambda : None, lambda : {'choices': get_available_loras()}, 'refresh-button')
+
     with gr.Accordion('Soft prompt', open=False):
         with gr.Row():
             shared.gradio['softprompts_menu'] = gr.Dropdown(choices=available_softprompts, value='None', label='Soft prompt')
@@ -245,11 +224,12 @@ def create_settings_menus(default_preset):
         with gr.Row():
             shared.gradio['upload_softprompt'] = gr.File(type='binary', file_types=['.zip'])
 
+    shared.gradio['model_menu'].change(load_model_wrapper, shared.gradio['model_menu'], shared.gradio['model_menu'], show_progress=True)
     shared.gradio['preset_menu'].change(load_preset_values, [shared.gradio[k] for k in ['preset_menu', 'generate_state']], [shared.gradio[k] for k in ['generate_state', 'do_sample', 'temperature', 'top_p', 'typical_p', 'repetition_penalty', 'encoder_repetition_penalty', 'top_k', 'min_length', 'no_repeat_ngram_size', 'num_beams', 'penalty_alpha', 'length_penalty', 'early_stopping']])
+    shared.gradio['lora_menu'].change(load_lora_wrapper, shared.gradio['lora_menu'], shared.gradio['lora_menu'], show_progress=True)
     shared.gradio['softprompts_menu'].change(load_soft_prompt, shared.gradio['softprompts_menu'], shared.gradio['softprompts_menu'], show_progress=True)
     shared.gradio['upload_softprompt'].upload(upload_soft_prompt, shared.gradio['upload_softprompt'], shared.gradio['softprompts_menu'])
 
-
 def set_interface_arguments(interface_mode, extensions, bool_active):
     modes = ["default", "notebook", "chat", "cai_chat"]
     cmd_list = vars(shared.args)
@@ -268,7 +248,6 @@ def set_interface_arguments(interface_mode, extensions, bool_active):
 
     shared.need_restart = True
 
-
 available_models = get_available_models()
 available_presets = get_available_presets()
 available_characters = get_available_characters()
@@ -317,8 +296,8 @@ else:
     default_text = load_prompt(shared.settings['prompts'][next((k for k in shared.settings['prompts'] if re.match(k.lower(), shared.model_name.lower())), 'default')])
 title ='Text generation web UI'
 
-
 def create_interface():
+
     gen_events = []
     if shared.args.extensions is not None and len(shared.args.extensions) > 0:
         extensions_module.load_extensions()
@@ -520,9 +499,6 @@ def create_interface():
             shared.gradio['Stop'].click(stop_everything_event, [], [], queue=False, cancels=gen_events if shared.args.no_stream else None)
             shared.gradio['interface'].load(None, None, None, _js=f"() => {{{ui.main_js}}}")
 
-        with gr.Tab("Model", elem_id="model-tab"):
-            create_model_menus()
-
         with gr.Tab("Training", elem_id="training-tab"):
             training.create_train_interface()
 
@@ -536,6 +512,7 @@ def create_interface():
             cmd_list = vars(shared.args)
             bool_list = [k for k in cmd_list if type(cmd_list[k]) is bool and k not in modes]
             bool_active = [k for k in bool_list if vars(shared.args)[k]]
+            #int_list = [k for k in cmd_list if type(k) is int]
 
             gr.Markdown("*Experimental*")
             shared.gradio['interface_modes_menu'] = gr.Dropdown(choices=modes, value=current_mode, label="Mode")
@@ -580,7 +557,6 @@ def create_interface():
     else:
         shared.gradio['interface'].launch(prevent_thread_lock=True, share=shared.args.share, server_port=shared.args.listen_port, inbrowser=shared.args.auto_launch, auth=auth)
 
-
 create_interface()
 
 while True: