Add PyPI-backed Samudra 2 Colab quickstart

README.md

@@ -6,13 +6,18 @@ SPDX-License-Identifier: CC-BY-4.0
 
 # Samudra
 
+[![Open Quickstart in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/m2lines/Samudra/blob/main/notebooks/quickstart.ipynb)
+
 [![Pre-commit](https://github.com/m2lines/Samudra/actions/workflows/pre-commit.yml/badge.svg)](https://github.com/m2lines/Samudra/actions/workflows/pre-commit.yml)
 [![Test CPU](https://github.com/m2lines/Samudra/actions/workflows/test.yml/badge.svg)](https://github.com/m2lines/Samudra/actions/workflows/test.yml)
 [![Benchmark CPU](https://github.com/m2lines/Samudra/actions/workflows/benchmarks.yml/badge.svg)](https://github.com/m2lines/Samudra/actions/workflows/benchmarks.yml)
 [Benchmark Results](https://m2lines.github.io/Samudra/dev/bench/)
 
 Samudra is a global ocean emulator described in ["Samudra: An AI Global Ocean Emulator for Climate"](https://arxiv.org/abs/2412.03795) and updated in ["Samudra 2: Scaling Ocean Emulators across Resolutions"](https://m2lines.github.io/Samudra/samudra2/). Samudra efficiently emulates the ocean component of a state-of-the-art climate model, accurately reproducing key ocean variables including sea surface height, horizontal velocities, temperature, and salinity, across their full depth.
 
+To try the complete Samudra 2 training workflow in a browser without HPC access or cloud
+credentials, open the [free-tier Colab quickstart](https://colab.research.google.com/github/m2lines/Samudra/blob/main/notebooks/quickstart.ipynb).
+
 ![Sea Surface Temperature of the Tropical Pacific: Ground Truth vs Samudra v2](https://raw.githubusercontent.com/m2lines/Samudra/main/docs/static/assets/sst_tropical_pacific_ultra_small.gif "Sea Surface Temperature of the Tropical Pacific: Ground Truth vs Samudra v2")
 
 > Ground truth (left) vs. Samudra 2 prediction (right) for sea surface temperature in the tropical Pacific.

docs/getting-started/quick-start.md

@@ -6,6 +6,15 @@ SPDX-License-Identifier: CC-BY-4.0
 
 # Quick Start
 
+## Run in Google Colab
+
+The [Colab quickstart](https://colab.research.google.com/github/m2lines/Samudra/blob/main/notebooks/quickstart.ipynb)
+installs the latest stable Samudra release from PyPI and trains a Samudra 2
+model on public 2° OM4 data using a free-tier GPU runtime. It streams the data
+directly from S3, so it requires no local installation, bulk data download, HPC
+access, or cloud credentials. The notebook is an onboarding smoke test rather
+than a scientifically useful training run.
+
 ## Training a Model
 
 Training is configured via YAML files. To launch a training run with the default Samudra configuration: