Assignment_04

Links:

• Airflow - UI Deployed on Google Compute Engine
• Streamlit - UI Deployed on Streamlit Cloud

---

Architecture:

---

Part 1: Automating Benchmarking of datasets

• DataPerf Source Source code and dataset stored as a branch dataperf_src

• Airflow DAG

  • clean_dir - Clean's the working directory
  • download_code - Download's the source code along with dataset from dataperf_src branch
  • generate_template - Generates the task_setup.yml file from the template.yml file replacing the user input.
  • run_python_script - Runs the computation script namely create_baselines.py, main.py and plotter.py
  • send_report_email - Send an email along with the generated report image
  • send_failure_email - Send an email informing the failure of job
  • send_ack_email - Send an email acknowledging the job run and for for further update email

Screenshots:

1. Airflow DAG's

2. Template file - link

paths:
embedding_folder: embeddings/
groundtruth_folder: data/
submission_folder: submissions/
results_folder: results/

tasks:
- data_id: {{ data_id }}
    train_size:  {{ train_size }}
    noise_level:  {{ noise_level }}
    test_size:  {{ test_size }}
    val_size:  {{ val_size }}

baselines:
    {% for element in  baselines %}
    - name: {{ element }}
    {% endfor %}

3. Streamlit UI

4. Job Run Acknowledgement Email

5. Job Run Report Email

6. Job Run Failure Email

7. Report Sample

---

Part 2: Technology evaluation

Snoopy Observation recorded here

---

Git Branches:

• airflow - Airflow Setup and DAG
• dataperf_src - DataPerf Source Code and Dataset
• Snoopy - Snoopy observation
• Streamlit - Streamlit User Interface