/cse252d-sp22-hw2

Primary LanguagePython

CSE252D Spring 2022 HW2

Instructions

  1. Attempt all questions.
  2. Please comment your code adequately.
  3. Include all relevant information such as text answers, output images in the Jupyter notebook.
  4. Academic integrity: The homework must be completed individually.
  5. Please select pages for each answer on Gradescope to facilitate smooth grading.
  6. Due date: Mon May 09, 11:59PM PDT.
  7. Access the HW by cloning this repository using,
    git clone https://github.com/ViLab-UCSD/cse252d-sp22-hw2.git
  8. Follow the Jupyter Notebook hw2_questions.ipynb.
  9. Follow the rest of README (this file) for instructions on how to setup your environment, data and compute.
  10. Submit the PDF version of your notebook and your code on Gradescope.
    (1) Convert the ipynb file to pdf and upload it to Homework 2 writeup. Select pages for each answer. Only include images and text relevant to your answers in this file.
    (2) Compress any supporting documents you think is necessary to justify your reasoning and answers into a zip file and upload it to Homework 2 code. Do not include any dataset, model or large data files.
  11. Rename your submission files as Lastname_Firstname.pdf and Lastname_Firstname.zip.

Frequently Asked Questions (FAQ)

I don't have access to a GPU. What are my options?
You can use the Data Science & Machine Learning Platform (DSMLP). It is a Kubernetes cluster with well defined usage policies and an easy-to-use interface. You can request for ephemeral pods/containers with a persistent storage space (your home directory). There are two ways to do this:

  1. Use the GUI at Datahub.
  2. Use a deployment script. Follow these intructions from HW1.

Where can I get the data?
The /datasets/cs252d-sp22-a00-public/hw2_data/ directory on DSMLP contains the required data for this HW. If you are using the cluster, change the paths in the Jupyter notebook such that they point to the correct directories. If you are using your own machine, you can use scp or rsync to download the data. Alternatively, you can also access it from GDrive (here and here).

How do I monitor my training progress on the cluster?
You can use tensorboard. Follow the instructions on this Piazza post on how to do this. Alternatively, we provide an easy to use bash script:

  1. Remote: /datasets/cs252d-sp22-a00-public/tb.sh {logdir} {port-number}.
  2. Local: ssh -N -L localhost:{port-number}:127.0.0.1:{port-number} {user}@dsmlp-login.ucsd.edu.
  3. Navigate to localhost:{port-number} in your browser.

How do I run my pods in background?
You can use the -b flag when you use the launch script. If you want to launch Jupyter sessions in background, use /datasets/cs252d-sp22-a00-public/cs252d-scipy-ml.sh instead of launch-scipy-ml.sh. Alternatively, use tmux.

module not found. How can I install the dependencies?
Use requirements.txt to install the dependencies
pip install -r requirements.txt.

More to come. We will update this section as and when we receive questions on Piazza.