Before running the script, make sure the requirements are met by installing from requirements.txt using
pip install -r requirements.txt

1. Configure the config.gin file in /configs

2. Use one of the following commands to perform training/evaluation/visualization. Make sure training is performed before evaluation or visualization. The experiment name which is used for training should be used for evaluation and visualization.

Note: Information on using checkpoints for evaluation and visualization can be found in the config file.

python main.py --train True --dir_name "Experiment-name"

python main.py --eval True --dir_name "Experiment-name"

python main.py --visualize True --dir_name "Experiment-name"

1. The files related to the experiment can be found under /experiments/Experiment-name/version_0

dl-lab-21w-team11
|-- architecture/
|   |-- base_model.py
|   |-- models.py
|-- configs/
|   |-- config.gin
|-- datasets/
|-- experiments/
|-- input_pipeline/
|   |-- data_loader.py
|-- utils/
|   |-- utils_misc.py
|   |-- utils_params.py
|-- .gitignore
|-- README.md
|-- requirements.txt
|-- har_batch.sh
|-- main.py
|-- train.py
|-- wandb-tune.py

• Hyperparameter Tuning:

Best Parameters for each model based on hyperparameter tuning

• Evaluation Accuracy:

  Model	Evaluation accuracy
  Simple LSTM	92.7
  Bidirectional LSTM	93.79
  MaxPool LSTM	93.86
  MaxPool Bidirectional LSTM	94.3

• Confusion Matrix: Maxpool Bidirectional LSTM