Project Setup Instruction

1. Setup and Activate virtual environment (name of environment - env)

   • Ubuntu: https://linuxize.com/post/how-to-create-python-virtual-environments-on-ubuntu-18-04/

2. Install requirements and Execute script files

   a) pip install -r requirements.txt
   b) sh scripts/script.sh
   c) sh scripts/glove.sh
   

3. Execute project setup File

   python setup.py develop
   

Project Execution Instruction

Module:

**dataset_generate.py** - 
   a) Write the module to Create the training dataset (train.csv)

**dataset.py** - 
   a) Create Tensor from training dataset [set the filename of training dataset (train.csv  in constants/constant.json under ***dataset_config.dataset_file*** key.
   b) Save the LabelEncoder for label to one-hot vector & vice-versa mapping in *data* folder
   
 **model.py** - 
 a) The module responsible for defining Neural Network Architecture, its error and optimization function
       
 **train.py** - 
 a) The module responsible for training the model for action item classification
 b) Once all the dataset and configuration (set the configuration in **constants/constants.json** file) run the following for training the model - ***python module/train.py***
 c) Once the training is over set the best model filename in *constants/constant.json* file under **predict_config.best_model_name** key for prediction with model
 d) Use following command from **project** folder to view training and validation loos & metrics:
         tensorboard --logdir=runs
 
 
 **predict.py**
 a) Pass list on input to predict the output label
         

Miscellaneous:

**utils** Folder - 
   a) **utils.py** - It contains the utility methods related to data-preprocessing, vectorization, finding evaluation metric, etc.
   b) **constant_reader.py** - It is used to read the hyper-parameter and other constants 
   
  **data** Folder
  b) **target.pkl** - Label Encoder for encoding and Decoding Label Name to and from its one-hot representation