add sections

docs/cellxgene.ipynb

@@ -1009,6 +1009,14 @@
     "## Train ML models"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "0cf26bfb",
+   "metadata": {},
+   "source": [
+    "### Create a dataloader from a `tiledbsoma` store"
+   ]
+  },
   {
    "cell_type": "markdown",
    "id": "9206d817",
@@ -1057,12 +1065,52 @@
   },
   {
    "cell_type": "markdown",
-   "id": "73b06fe6",
+   "id": "8884cb2c",
+   "metadata": {},
+   "source": [
+    "### Create a dataloader from a `Collection` of `AnnData` objects"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "79cacdba",
+   "metadata": {},
+   "source": [
+    "You can also directly train ML models on very large collections of `AnnData` objects."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "04c41eae",
    "metadata": {},
    "source": [
-    "You can also directly train ML models on very large collections of AnnData objects.\n",
+    "{meth}`~lamindb.Collection.mapped` caches `AnnData` objects to disk and creates a map-style dataset that performs a virtual join of the features of the underlying `AnnData` objects."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "bcf4e905",
+   "metadata": {},
+   "source": [
+    "```\n",
+    "from torch.utils.data import DataLoader\n",
+    "\n",
+    "collection = ln.Collection.filter(name=\"cellxgene-census\", version=census_version).one()\n",
+    "\n",
+    "dataset = collection.mapped(obs_keys=[features.cell_type], join=\"outer\")\n",
+    "dataloader = DataLoader(dataset, batch_size=128, shuffle=True)\n",
     "\n",
-    "See {doc}`docs:scrna5`."
+    "for batch in dataloader:\n",
+    "    pass\n",
+    "```"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "73b06fe6",
+   "metadata": {},
+   "source": [
+    "For further details see {doc}`docs:scrna5`."
    ]
   },
   {