Hyperscale ES Lab

This demo follows the actual structure of the repo: environment check, data prep, baseline, reproduction, ablation, and artifact folders. It is less about running the full paper online and more about exposing the experiment workflow clearly.

What this demo focuses on

1. The actual notebook sequence defined in the repo.

2. The data and artifact layout used across the experiment workflow.

3. A clearer story for visitors who want to understand how the lab is structured.

Why not run the full paper reproduction in-browser?

The repo is notebook-first and aimed at guided experimentation.

A website demo should explain the workflow honestly instead of faking heavy experiment execution.

This gives a cleaner path for a future notebook-backed or sandbox-backed demo later.

Notebook stages

Notebook-like demo

Sandbox-ready direction

00_env_check.ipynb

Verify the Python environment and package setup before running experiments.

Package readiness

Notebook runtime check

Notebook cells

Notebook note

Ready

This stage checks whether the baseline scientific Python stack is available and whether the notebook runtime is wired correctly.

Cell 1

Ready

import numpy as np
import pandas as pd
import matplotlib

print('Environment looks ready')
print('numpy:', np.__version__)
print('pandas:', pd.__version__)
print('matplotlib:', matplotlib.__version__)

Preview output

Environment looks ready
numpy: 2.x.x
pandas: 2.x.x
matplotlib: 3.x.x

Expected outcome

Ready

The user should see a quick confirmation that core packages import successfully and the runtime is healthy.

Lab storage and artifact map

data/raw

Original inputs and source material used before preprocessing.

data/processed

Prepared datasets and transformed experiment inputs.

reports

Metrics, notes, experiment summaries, and writeups.

figures

Plots and visual outputs exported from notebooks.

src

Shared helper code used by notebooks and experiment steps.

Next step: wire selected code cells to a Vercel Sandbox executor so visitors can run short, predefined experiment cells instead of only viewing previews.

View repository