- Timothy Ta ([email protected])
- Ariel Chiang ([email protected])
- Jingtian Yao ([email protected])
Github Repo: https://github.com/lanceyjt/cs598-dlh-team72
This Github repo intends to reproduce the CASTER (Huang et al., 2020) neural network framework and experiments discussed in the original paper. It incorporates original code sourced from the GitHub repository: https://github.com/kexinhuang12345/CASTER/tree/master.
The Jupyter notebook DL4H_Team72.ipynb
contains all the information for this project, including model architecture, model training, evaluation, results and discussions.
Use one of the methods below to run the notebook.
The training process is implemented in an Amazon SageMaker notebook with an ml.p3.8xlarge instance. Click here for more information.
Follow the steps below to reproduce the code:
-
Follow the instructions to create an AWS account, or log in to an existing AWS account.
-
Follow the instructions to create a SageMaker instance, ml.p3.8xlarge instance is recommended, but can consider other instance types with GPUs too. (Note: You will pay for the instance computation and storage, click here for more pricing information)
-
After the status of the SageMaker instance becomes "Active", click the "Open JupyterLab" from the upper right.
-
Launch a terminal window from the instance, run the commands below:
cd ~/SageMaker
git clone https://github.com/lanceyjt/cs598-dlh-team72
- Run the code blocks in the notebook
DL4H_Team72.ipynb
in sequential order. The parameterLOAD_FINAL_MODEL
in the first block is set to True by default. If you would like to perform the model training process, set it to False.