feat: adding YOLONAS to ml backend examples

label_studio_ml/examples/yolonas/.env

@@ -0,0 +1,12 @@
+CHECKPOINT_FILE="/home/testuser/app/model.pth"
+PORT=9090
+YOLO_LABELS=/home/testuser/app/labels.txt
+IOU_THRESHOLD=0.25
+SCORE_THRESHOLD=0.4
+IMG_SIZE=1280
+DEVICE=cpu
+ENDPOINT_URL=<specify minio address http://myminio:9000>
+AWS_ACCESS_KEY_ID=minio
+AWS_SECRET_ACCESS_KEY=<specify minio password>
+LABEL_STUDIO_HOSTNAME=<specify label studio address with port like http://mylabelstudio.com:8080 >
+YOLO_MODEL_TYPE=yolo_nas_m

label_studio_ml/examples/yolonas/Dockerfile

@@ -0,0 +1,5 @@
+FROM bodbe/yolonas
+USER testuser
+WORKDIR /home/testuser/app
+COPY _wsgi.py ./
+COPY model.py ./

label_studio_ml/examples/yolonas/Dockerfile.full

@@ -0,0 +1,63 @@
+FROM nvidia/cuda:11.2.2-cudnn8-devel-ubuntu18.04
+
+ENV TZ 'Europe/Moscow'
+RUN echo $TZ > /etc/timezone
+
+RUN apt-get update && apt-get install -y locales sudo
+RUN sed -i -e 's/# ru_RU.UTF-8 UTF-8/ru_RU.UTF-8 UTF-8/' /etc/locale.gen && \
+    dpkg-reconfigure --frontend=noninteractive locales && \
+    update-locale LANG=ru_RU.UTF-8
+ENV LANG ru_RU.UTF-8
+ENV LANGUAGE ru_RU
+ENV LC_ALL ru_RU.UTF-8
+
+RUN apt-get update && apt-get install -y python3.8 python3.8-dev python3-pip git
+#RUN apt-get upgrade python3-pip
+RUN update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.8 1
+
+RUN apt-get install -y ffmpeg libsm6 libxext6 libgl1-mesa-glx libgl1
+RUN apt-get install -y wget nano sudo
+RUN wget https://bootstrap.pypa.io/get-pip.py && python3.8 get-pip.py
+RUN apt-get clean
+
+RUN useradd -ms /bin/bash testuser && \
+    echo "testuser:testuser" | chpasswd && \
+    usermod -aG sudo testuser && \
+    chmod 777 -R /root
+
+WORKDIR /home/testuser
+
+USER testuser
+ENV PATH="/home/testuser/.local/bin:${PATH}"
+
+RUN mkdir -p /home/testuser/.jupyter/lab/user-settings/@jupyterlab/terminal-extension
+COPY --chown=testuser plugin.jupyterlab-settings /home/testuser/.jupyter/lab/user-settings/@jupyterlab/terminal-extension/
+COPY --chown=testuser jupyter_lab_config.py /home/testuser/.jupyter/
+
+RUN openssl req -x509 -sha256 -nodes -days 3650 -newkey rsa:4096 -keyout .jupyter/jupyter.key -out .jupyter/jupyter.pem \
+ -subj "/C=RU/ST=Uranopolis/L=Karyes/O=Space/OU=DS/CN=agion.oros"
+
+RUN python3.8 -m pip install nvidia-pyindex
+RUN python3.8 -m pip install pytorch-quantization==2.1.2
+COPY requirements.txt .
+RUN python3.8 -m pip install -U -r requirements.txt
+RUN python3.8 -m pip install super-gradients==3.1.2
+
+############
+## add-ons for late package installs
+############
+#RUN python3.8 -m pip install redis rq label-studio-ml
+
+# setting up token string
+# you must run the build with command
+# docker build --build-arg token_string=<whatever your token is> . -t torch_custom
+ARG token_string
+ENV JUPYTER_TOKEN=${token_string}
+
+ENV SHELL="/bin/bash"
+
+RUN mkdir /home/testuser/.clearml && mkdir /home/testuser/.ssh
+RUN ln -s /usr/bin/python3 ~/.local/bin/python
+WORKDIR /home/testuser
+
+CMD ["jupyter", "lab"] 

label_studio_ml/examples/yolonas/README.md

@@ -0,0 +1,46 @@
+# YOLONAS ML Backend for Label Studio  
+
+## Intro  
+Use Deci AI [YOLONAS](https://github.com/Deci-AI/super-gradients/blob/master/YOLONAS.md) model with Label Studio.  
+
+## Setup
+### 0. Important things to note  
+ - This ML backend is designed to work in **docker**. You can run in on host but this manual does not cover that.  
+ - There is no easy way to run ML backend with GPU support - you get an error `RuntimeError: Cannot re-initialize CUDA in forked subprocess. To use CUDA with multiprocessing, you must use the 'spawn' start method`
+ - Single image inference on CPU takes about a second.
+ - Base docker image has a `jupyter lab` command. So you can comment a `command` in `docker-compose.yml`, uncomment `8888:8888` for port mapping and use it as jupyter lab with password `change-me` over https.
+ - Main tested scenario has been for s3 cloud storage with custom endpoint url. Other storage options are not guaranteed to work. 
+### 1. Clone this repo
+### 2. Get model weights  
+### 3. Adjust variables  
+Adjust these variables in `.env` file.
+```
+CHECKPOINT_FILE="/home/testuser/app/model.pth"
+PORT=9090
+YOLO_LABELS=/home/testuser/app/labels.txt
+IOU_THRESHOLD=0.25
+SCORE_THRESHOLD=0.4
+IMG_SIZE=1280
+DEVICE=cpu
+ENDPOINT_URL=<specify minio address http://myminio:9000>
+AWS_ACCESS_KEY_ID=minio
+AWS_SECRET_ACCESS_KEY=<specify minio password>
+LABEL_STUDIO_HOSTNAME=<specify label studio address with port like http://mylabelstudio.com:8080 >
+YOLO_MODEL_TYPE=yolo_nas_m
+```  
+
+`YOLO_LABELS=/home/testuser/app/labels.txt` file with labels - each label on new line.  
+Labels should be the same in labeling interface and in this file. 
+If yolo labels differ you need to provide `LABELS_FILE` variable with mapping from Label studio label to yolo label like  `{"airplane": "Boeing"}`
+
+### 4. Build docker image  
+Run `docker compose build` to build an image.  
+Base image `bodbe/yolonas` is built with [Dockerfile.full](Dockerfile.full).  
+
+### 5. Run ML Backend  
+Run `docker compose up -d` 
+
+### 6. How to run on GPU  
+ - Update DEVICE variable in `.env` file to `cuda:0`
+ - Uncomment `deploy` section in `docker-compose.yml`
+ - Change `command` section in `docker-compose.yml` to `bash -c "python _wsgi.py"

label_studio_ml/examples/yolonas/_wsgi.py

@@ -0,0 +1,126 @@
+import os
+import argparse
+import logging
+import logging.config
+
+logging.config.dictConfig({
+  "version": 1,
+  "formatters": {
+    "standard": {
+      "format": "[%(asctime)s] [%(levelname)s] [%(name)s::%(funcName)s::%(lineno)d] %(message)s"
+    }
+  },
+  "handlers": {
+    "console": {
+      "class": "logging.StreamHandler",
+      "level": "DEBUG",
+      "stream": "ext://sys.stdout",
+      "formatter": "standard"
+    }
+  },
+  "root": {
+    "level": "ERROR",
+    "handlers": [
+      "console"
+    ],
+    "propagate": True
+  }
+})
+
+from label_studio_ml.api import init_app
+from model import ObjectDetectorModel
+
+
+_DEFAULT_CONFIG_PATH = os.path.join(os.path.dirname(__file__), 'config.json')
+
+
+def get_kwargs_from_config(config_path=_DEFAULT_CONFIG_PATH):
+    if not os.path.exists(config_path):
+        return dict()
+    with open(config_path) as f:
+        config = json.load(f)
+    assert isinstance(config, dict)
+    return config
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description='Label studio')
+    parser.add_argument(
+        '-p', '--port', dest='port', type=int, default=9090,
+        help='Server port')
+    parser.add_argument(
+        '--host', dest='host', type=str, default='0.0.0.0',
+        help='Server host')
+    parser.add_argument(
+        '--kwargs', '--with', dest='kwargs', metavar='KEY=VAL', nargs='+', type=lambda kv: kv.split('='),
+        help='Additional LabelStudioMLBase model initialization kwargs')
+    parser.add_argument(
+        '-d', '--debug', dest='debug', action='store_true',
+        help='Switch debug mode')
+    parser.add_argument(
+        '--log-level', dest='log_level', choices=['DEBUG', 'INFO', 'WARNING', 'ERROR'], default=None,
+        help='Logging level')
+    parser.add_argument(
+        '--model-dir', dest='model_dir', default=os.path.dirname(__file__),
+        help='Directory where models are stored (relative to the project directory)')
+    parser.add_argument(
+        '--check', dest='check', action='store_true',
+        help='Validate model instance before launching server')
+
+    args = parser.parse_args()
+
+    # setup logging level
+    if args.log_level:
+        logging.root.setLevel(args.log_level)
+
+    def isfloat(value):
+        try:
+            float(value)
+            return True
+        except ValueError:
+            return False
+
+    def parse_kwargs():
+        param = dict()
+        for k, v in args.kwargs:
+            if v.isdigit():
+                param[k] = int(v)
+            elif v == 'True' or v == 'true':
+                param[k] = True
+            elif v == 'False' or v == 'False':
+                param[k] = False
+            elif isfloat(v):
+                param[k] = float(v)
+            else:
+                param[k] = v
+        return param
+
+    kwargs = get_kwargs_from_config()
+
+    if args.kwargs:
+        kwargs.update(parse_kwargs())
+
+    if args.check:
+        print('Check "' + ObjectDetectorModel.__name__ + '" instance creation..')
+        model = ObjectDetectorModel(**kwargs)
+
+    app = init_app(
+        model_class=ObjectDetectorModel,
+        model_dir=os.environ.get('MODEL_DIR', args.model_dir),
+        redis_queue=os.environ.get('RQ_QUEUE_NAME', 'default'),
+        redis_host=os.environ.get('REDIS_HOST', 'localhost'),
+        redis_port=os.environ.get('REDIS_PORT', 6379),
+        **kwargs
+    )
+
+    app.run(host=args.host, port=args.port, debug=args.debug)
+
+else:
+    # for uWSGI use
+    app = init_app(
+        model_class=ObjectDetectorModel,
+        model_dir=os.environ.get('MODEL_DIR', os.path.dirname(__file__)),
+        redis_queue=os.environ.get('RQ_QUEUE_NAME', 'default'),
+        redis_host=os.environ.get('REDIS_HOST', 'localhost'),
+        redis_port=os.environ.get('REDIS_PORT', 6379)
+    )

label_studio_ml/examples/yolonas/docker-compose.yml

@@ -0,0 +1,57 @@
+version: "3.8"
+
+services:
+  redis:
+    image: redis:alpine
+    container_name: redis
+    hostname: redis
+    volumes:
+      - "./data/redis:/data"
+    expose:
+      - 6379
+  server:
+    build: .
+    container_name: server
+    environment:
+#      file with checkpoint from yolo nas
+      - CHECKPOINT_FILE=${CHECKPOINT_FILE:-/home/testuser/app/model.pth}
+#      file with labels - each label on new line
+#      labels should be the same in labeling interface and in this file
+#      otherwise need to provide LABELS_FILE variable with mapping from
+#      Label studio label to Yolo label like  {"airplane": "Boeing"}
+      - YOLO_LABELS=${YOLO_LABELS:-/home/testuser/app/labels.txt}
+      - LABELS_FILE=${LABELS_FILE:-}
+      - IOU_THRESHOLD=${IOU_THRESHOLD:-0.25}
+      - SCORE_THRESHOLD=${SCORE_THRESHOLD:-0.45}
+#     target resolution to feed to yolo
+      - IMG_SIZE=${IMG_SIZE:-1280}
+      - DEVICE=${DEVICE:-cpu}
+#      custom endpoint url for s3 storage that holds files in Label Studio
+      - ENDPOINT_URL=${ENDPOINT_URL:-}
+      - AWS_ACCESS_KEY_ID=${AWS_ACCESS_KEY_ID:-}
+      - AWS_SECRET_ACCESS_KEY=${AWS_SECRET_ACCESS_KEY}
+      - LABEL_STUDIO_HOSTNAME=${LABEL_STUDIO_HOSTNAME}
+#      yolo model specification
+      - YOLO_MODEL_TYPE=${YOLO_MODEL_TYPE:-yolo_nas_m}
+      - RQ_QUEUE_NAME=default
+      - REDIS_HOST=redis
+      - REDIS_PORT=6379
+      - MODEL_DIR=./
+    ports:
+# uncomment to use jupyter lab
+#      - 8888:8888
+      - ${PORT}:${PORT:-9090}
+    restart: always
+    volumes:
+      - ./model.pth:${CHECKPOINT_FILE}
+      - ./labels.txt:${YOLO_LABELS}
+    command: bash -c "exec gunicorn --preload --bind :${PORT} --workers 1 --threads 8 --timeout 0 _wsgi:app"
+# uncomment if you wand a GPU access inside your docker
+#    command: bash -c "python _wsgi.py"
+#    deploy:
+#      resources:
+#        reservations:
+#          devices:
+#          - driver: nvidia
+#            device_ids: [ '0']
+#            capabilities: [ gpu ]