train.py

import requests
import json

try:
    from utils.logger import get_logger

    logger = get_logger()
except ImportError:
    import logging
    logger = logging.getLogger(__name__)

API_UNAVAILABLE = False




nlu_files = [
    'data/user-data/user_data.md',
    'data/user-data/user_data_2.md',
    'data/manually-generated/commands.md',
    'data/manually-generated/conversation.md',
    'data/auto-generated/commands.md',
    'data/auto-generated/clarification.md',
    'data/rasa-interactive/nlu.md'
    ]

stories_files = [
    'data/rasa-interactive/stories.md',
    'data/manually-generated/stories.md'
    ]


def train_model():
    try:
        # get the domain file from the API
        headers_domain = {'Accept': "Accept: application/json"}
        domain_request = requests.get("http://localhost:5005/domain", headers=headers_domain)
        print(domain_request.status_code)

        # load the NLU and stories data from training files
        with open('pipelines/supervised_embeddings.yml', 'r') as f:
            config = f.read()
    except Exception as e:
        logger.warning(e)
        API_UNAVAILABLE = True

    if API_UNAVAILABLE:
        logger.warning("API unavailable, didn't re-train model")
        return

    try:
        nlu_data = ''
        for file in nlu_files:
            with open(file, 'r') as f:
                nlu_data += f.read()+'\n'

        stories_data = ''
        for file in stories_files:
            with open(file, 'r') as f:
                stories_data += f.read()+'\n'

        # construct a training request
        train_request = {
            "domain": domain_request.text,
            "config": config,
            "nlu": nlu_data,
            "stories": stories_data,
            "force": False,
            "save_to_default_model_directory": False
        }

        headers_train = {
            'Content-Type': "application/json",
            "Accept": "*/*",
            "Accept-Encoding": "gzip, deflate, br"
        }
        train_request = requests.post("http://localhost:5005/model/train",
                                      headers=headers_train,
                                      data=json.dumps(train_request))

        print(train_request.status_code)

        if train_request.status_code == 200:
            # write model to file
            with open('models/test.tar.gz','wb') as f:
                train_request.raw.decode_content = True
                f.write(train_request.content)

            # load the model
            model_request = {
              "model_file": "models/test.tar.gz"
            }

            headers_model = {
                'Accept': "*/*"
            }

            load_request = requests.put("http://localhost:5005/model",
                                         headers=headers_model,
                                         data=json.dumps(model_request))

            print(load_request.status_code)
        else:
            print(train_request.text[:300])
    except Exception as e:
        logger.warning(e)


if __name__ == "__main__":
    train_model()