Root
└── data 
│   ├──  (all datasets go to this folder)
└── pipeline
│   ├──  DialogueStateManager.py
│   ├──  EmotionClassifier.py
│   ├──  IntentClassifier.py
│   └──  ResponseGenerator.py
├── models
│   ├── emotions
│   │	└── Emotion classification model
│   └── intents
│   │	└── Intent classification model
│   └── gpt2
│	└── response generator model
├── pipeline.py (main of the dialogue system)
├── train_emotion.py (for training emotion classifier)
├── train_intent.py (for training intention classifier)
├── train_gpt2.py (for training response generator)
├── check_maxlength.py (for checking maxlength of final prompt)
└── dataset.py (dataset consolidated for all trainings)

Dataset links please refer to links in Final Report submitted.

Models link: link

1. Install Anaconda in the machine and create environment, make sure CUDA and GPU are properly installed if you want to use GPU for finetuning or inference in pipeline. conda create --name <env_name> python=3.9

2. Download the repo from Github and all the models from cloud drive

3. Replace all 3 paths to the model directories for 3 models in pipeline> load_models()

4. Replace the path to the persona dataset to get the persona to where it is stored.

5. Go to the root directory of the system cd <path_to_root_dir_of_repo>

6. Install all necessary requirements pip install -r requirements.txt python -m spacy download en_core_web_sm

7. Run the system python pipeline.py After the chatbot is prepared as show in console, start talking with the chatbot.

8. Anytime you want to end the conversation type exit in the conversation to signal the system for ending the process.