docs(notebook): add show4dstem_colab.ipynb for one-click Colab workshop

notebooks/show4dstem_colab.ipynb

@@ -0,0 +1,145 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "b5f2091f",
+   "metadata": {},
+   "source": [
+    "# Workshop: Show4DSTEM from Hugging Face via `Dataset4dstem.from_tensor` (no local install needed)\n",
+    "\n",
+    "This notebook runs in **Google Colab** (or any Jupyter with internet + Chrome). You need\n",
+    "only `quantem.widget` (TestPyPI rc) and `quantem` (dev fork; the `from_tensor`\n",
+    "constructor lives on the `dataset-support-torch` branch). No CUDA, no h5, no\n",
+    "`quantem.live`.\n",
+    "\n",
+    "[Open in Colab](https://colab.research.google.com/?url=https%3A%2F%2Fhuggingface.co%2Fdatasets%2Fbobleesj%2Fquantem-data%2Fresolve%2Fmain%2Fnotebooks%2Fshow4dstem_colab.ipynb)\n",
+    "\n",
+    "Flow: download a pre-binned NumPy 4D-STEM dataset from Hugging Face -> wrap it as a\n",
+    "torch tensor -> `Dataset4dstem.from_tensor` carries the sampling -> `Show4DSTEM` renders\n",
+    "in your browser via WebGPU. Drag the scan cursor; the diffraction pattern updates live."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "d9055753",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# quantem.widget rc from TestPyPI; quantem from the dev fork branch that\n",
+    "# carries Dataset4dstem.from_tensor (the torch-native constructor).\n",
+    "!pip install -q --pre --extra-index-url https://test.pypi.org/simple/ quantem.widget huggingface_hub\n",
+    "!pip install -q git+https://github.com/bobleesj/quantem.git@dataset-support-torch"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "95dc87e5",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import quantem, quantem.widget\n",
+    "print(\"quantem        \", quantem.__version__)\n",
+    "print(\"quantem.widget \", quantem.widget.__version__)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "590f83f1",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Download the pre-binned NumPy 4D-STEM dataset from Hugging Face. Public,\n",
+    "# cached locally after the first run. ~302 MB; takes 5-30 s on Colab.\n",
+    "import os, json\n",
+    "import numpy as np\n",
+    "from huggingface_hub import snapshot_download\n",
+    "\n",
+    "ASSET = \"gold_512_npy_bin8\"          # or \"gold_512_npy_bin4\" for the larger 1.2 GB version\n",
+    "folder = snapshot_download(\n",
+    "    \"bobleesj/quantem-data\",\n",
+    "    repo_type=\"dataset\",\n",
+    "    allow_patterns=[f\"4dstem/{ASSET}/*\"],\n",
+    ")\n",
+    "asset_dir = os.path.join(folder, \"4dstem\", ASSET)\n",
+    "data = np.load(os.path.join(asset_dir, \"data.npy\"))            # (512, 512, 24, 24) uint16\n",
+    "meta = json.load(open(os.path.join(asset_dir, \"meta.json\")))   # sampling + units + optics\n",
+    "\n",
+    "print(\"shape:\", data.shape, \"dtype:\", data.dtype)\n",
+    "print(\"sampling:\", meta[\"sampling\"], meta[\"units\"])\n",
+    "print(\"optics: voltage=%s kV, probe=%s mrad, CL=%s mm\" % (\n",
+    "    meta.get(\"voltage_kV\"), meta.get(\"probe_semiangle_mrad\"), meta.get(\"camera_length_mm\"),\n",
+    "))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "1e14a283",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# numpy -> torch tensor -> Dataset4dstem.from_tensor. The dataset carries\n",
+    "# `sampling` + `units` so the widget paints the scale bar and axis labels.\n",
+    "import torch\n",
+    "from quantem.core.datastructures import Dataset4dstem\n",
+    "\n",
+    "dset = Dataset4dstem.from_tensor(\n",
+    "    torch.from_numpy(data),\n",
+    "    sampling=meta[\"sampling\"],\n",
+    "    units=meta[\"units\"],\n",
+    "    name=meta[\"name\"],\n",
+    ")\n",
+    "print(\"dataset:\", type(dset).__name__, \"shape:\", dset.shape, \"sampling:\", dset.sampling)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "d99029a3",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Render in your browser via WebGPU. Drag the cursor in the real-space image on the\n",
+    "# left; the CBED on the right updates in real time.\n",
+    "from quantem.widget import Show4DSTEM\n",
+    "\n",
+    "Show4DSTEM(dset)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "c1d194fd",
+   "metadata": {},
+   "source": [
+    "## What just happened\n",
+    "\n",
+    "1. Two `pip install` lines: `quantem.widget` (rc from TestPyPI) + `quantem` (dev fork\n",
+    "   branch with the torch-native `Dataset4dstem.from_tensor`).\n",
+    "2. `snapshot_download` pulled a pre-binned NumPy file (~302 MB) from `bobleesj/quantem-data`.\n",
+    "3. `np.load` parsed it directly - no custom decompression, no h5 parser.\n",
+    "4. `Dataset4dstem.from_tensor` wrapped the torch tensor with sampling + units.\n",
+    "5. `Show4DSTEM` rendered it in your browser via WebGPU.\n",
+    "\n",
+    "No CUDA on Colab. No quantem.live. No private repos.\n",
+    "\n",
+    "## Try next\n",
+    "\n",
+    "- Swap `ASSET` to `\"gold_512_npy_bin4\"` for the sharper (larger, 1.2 GB) version.\n",
+    "- Try `gold_haadf_npy` with `Show2D` (a 2D HAADF image; same pattern, different widget).\n",
+    "- Bring your own 4D-STEM data: save as `(scan_row, scan_col, k_row, k_col)` NumPy +\n",
+    "  a `meta.json` carrying sampling + units, and pass through the same pipeline."
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "name": "python3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}