/stand_squad_ru

Primary LanguagePython

Demo stand. Model: SQuAD (Russian)

Installation and start

  1. Clone the repo and cd to project root: 
    git clone https://github.com/deepmipt/stand_squad_ru.git
cd stand_squad_ru
  2. Run script to download and unpack model components: 
    ./download_components.sh
  3. Create a virtual environment with Python 3.6: 
    virtualenv env -p python3.6
  4. Activate the environment: 
    source ./env/bin/activate
  5. Install requirements: 
    pip install -r requirements.txt
  6. Download NLTK data: 
    $ python3
>>> import nltk
>>> nltk.download('punkt')
  7. Specify model endpoint host (api_host) and port (api_port) in squad_agent_config.json
  8. Specify CUDA_VISIBLE_DEVICES and virtual environment path (if necessary) in run_ru_squad.sh
  9. Run model: 
    ./run_ru_squad.sh

Building and running with Docker:

  1. If necessary, build Base Docker image from:

    https://github.com/deepmipt/stand_docker_cuda

  2. Clone the repo and cd to project root:

    git clone https://github.com/deepmipt/stand_squad_ru.git
cd stand_squad_ru

  3. Build Docker image:

    sudo docker build -t stand/squad_ru .

  4. Run Docker image:

    sudo docker run -p <host_port>:6005 --runtime=nvidia --device=/dev/nvidia<gpu_unit_int_id> -v </path/to/host/vol/map/dir>:/logs stand/squad_ru