lincc-frameworks · Graciaaa3 · Jun 26, 2026 · Jun 25, 2026 · Jun 25, 2026 · Jun 25, 2026
diff --git a/docs/_static/diagrams/hyrax_schematic_excalidraw.excalidraw.png b/docs/_static/diagrams/hyrax_schematic_excalidraw.excalidraw.png
diff --git a/docs/common_workflows.rst b/docs/common_workflows.rst
@@ -18,7 +18,7 @@ Common workflows
    Checkpointing <pre_executed/hyrax_checkpointing>
    Loading pretrained weights <pre_executed/load_pretrained_model>
    Hyperparameter tuning <pre_executed/hyperparameter_tuning_examples>
-   Using UMAP <pre_executed/using_umap>
+   Using reduce_dimensions <pre_executed/using_reduce_dimensions>
    Custom training metrics <notebooks/custom_training_metrics>
    Dataset custom collation <notebooks/custom_dataset_collation>
    Customizing prepare_inputs <notebooks/custom_prepare_inputs>

diff --git a/docs/index.rst b/docs/index.rst
@@ -67,7 +67,7 @@ search, and interactive exploration.
    h.train()                 # Train any PyTorch model with automatic logging & multi-GPU support
    h.infer()                 # Run inference and store results
    h.save_to_database()      # Index embeddings in a vector database
-   h.umap()                  # Reduce latent vectors to 2D/3D with UMAP
+   h.reduce_dimensions()     # Reduce latent vectors to 2D/3D with dimensionality-reduction algorithm
    h.visualize()             # Interactively explore latent spaces in 2D or 3D
    db = h.database_connection()
    v = ...                   # numpy vector representing the object to search for

diff --git a/docs/pre_executed/unsupervised_hsc_full_pipeline.ipynb b/docs/pre_executed/unsupervised_hsc_full_pipeline.ipynb
@@ -220,20 +220,21 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "[2026-03-27 18:13:08,743 hyrax.verbs.umap:INFO] Saving UMAP results to /mmfs1/gscratch/dirac/aritrag/repos/hyrax/docs/pre_executed/results/20260327-181308-umap-QNSR\n",
-      "[2026-03-27 18:13:08,778 hyrax.verbs.umap:INFO] Fitting the UMAP\n",
-      "[2026-03-27 18:13:18,448 hyrax.verbs.umap:INFO] Saving fitted UMAP Reducer\n"
+      "[2026-06-24 17:04:27,154 hyrax.verbs.reduce_dimensions:INFO] Saving reduction results using umap to /Users/qiyuwang/Desktop/hyrax/docs/pre_executed/results/20260624-170427-umap-LNhc\n",
+      "[2026-06-24 17:04:27,159 hyrax.verbs.reduce_dimensions:INFO] No model_path specified. A new model will be fitted.\n",
+      "[2026-06-24 17:04:27,159 hyrax.verbs.reduction_algorithms.umap:INFO] Fitting the UMAP\n",
+      "[2026-06-24 17:04:31,013 hyrax.verbs.reduce_dimensions:INFO] Saving fitted umap reducer to result directory\n"
      ]
     },
     {
      "data": {
       "application/vnd.jupyter.widget-view+json": {
-       "model_id": "6f4d2b9c8e0b495ba1d76b0e44c55182",
+       "model_id": "f67593f48c554940a6671b3fdf129faa",
        "version_major": 2,
        "version_minor": 0
       },
       "text/plain": [
-       "Creating lower dimensional representation using UMAP::   0%|          | 0/2 [00:00<?, ?it/s]"
+       "Creating lower dimensional representation using UMAP::   0%|          | 0/1 [00:00<?, ?it/s]"
       ]
      },
      "metadata": {},
@@ -243,16 +244,15 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "\u001b[90m[\u001b[0m2026-03-28T01:13:21Z \u001b[33mWARN \u001b[0m lance::dataset::write::insert\u001b[90m]\u001b[0m No existing dataset at /mmfs1/gscratch/dirac/aritrag/repos/hyrax/docs/pre_executed/results/20260327-181308-umap-QNSR/lance_db/results.lance, it will be created\n",
-      "[2026-03-27 18:13:22,511 hyrax.datasets.result_dataset:INFO] Optimizing Lance table after 2 batches\n",
-      "[2026-03-27 18:13:22,530 hyrax.datasets.result_dataset:INFO] Lance table optimization complete\n",
-      "[2026-03-27 18:13:22,530 hyrax.verbs.umap:INFO] Finished transforming all data through UMAP\n"
+      "[2026-06-24 17:04:32,764 hyrax.datasets.result_dataset:INFO] Optimizing Lance table after 1 batches\n",
+      "[2026-06-24 17:04:32,767 hyrax.datasets.result_dataset:INFO] Lance table optimization complete\n",
+      "[2026-06-24 17:04:32,767 hyrax.verbs.reduce_dimensions:INFO] Finished transforming all data with umap\n"
      ]
     },
     {
      "data": {
       "text/plain": [
-       "<hyrax.datasets.result_dataset.ResultDataset at 0x7f04518f8070>"
+       "<hyrax.datasets.result_dataset.ResultDataset at 0x3d7e76120>"
       ]
      },
      "execution_count": 6,
@@ -261,8 +261,8 @@
     }
    ],
    "source": [
-    "h.config[\"umap\"][\"UMAP\"][\"n_components\"] = 2\n",
-    "h.umap()"
+    "h.set_config(\"reduce.umap.kwargs.n_components\", 2)\n",
+    "h.reduce_dimensions(algorithm=\"umap\")"
    ]
   },
   {
@@ -281,7 +281,7 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "[2026-03-27 18:13:22,602 hyrax.prepare:INFO] Finished Prepare\n"
+      "[2026-06-24 17:04:32,797 hyrax.verbs.prepare:INFO] Finished Prepare\n"
      ]
     }
    ],
@@ -1100,7 +1100,7 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python 3 (ipykernel)",
+   "display_name": "hyrax",
    "language": "python",
    "name": "python3"
   },
@@ -1114,7 +1114,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.10.16"
+   "version": "3.12.13"
   },
   "nbsphinx": {
    "execute": "never"

diff --git a/docs/pre_executed/unsupervised_image_extragalactic.ipynb b/docs/pre_executed/unsupervised_image_extragalactic.ipynb
diff --git a/docs/pre_executed/using_reduce_dimensions.ipynb b/docs/pre_executed/using_reduce_dimensions.ipynb
diff --git a/docs/pre_executed/using_umap.ipynb b/docs/pre_executed/using_umap.ipynb
diff --git a/docs/pre_executed/working_with_results_data.ipynb b/docs/pre_executed/working_with_results_data.ipynb
@@ -7,7 +7,7 @@
    "source": [
     "# Result data in Hyrax\n",
     "\n",
-    "Hyrax verbs like `infer` and `umap` return results as instances of `ResultDataset`, a convenience layer over the underlying [lance file format](https://lance.org/) stored on disk.\n",
+    "Hyrax verbs like `infer` and `reduce_dimensions` return results as instances of `ResultDataset`, a convenience layer over the underlying [lance file format](https://lance.org/) stored on disk.\n",
     "\n",
     "This notebook follows the [Getting Started](<getting_started.ipynb>) notebook, training a `HyraxCNN` model on CIFAR-10 data, then explores how to work directly with inference results."
    ]

diff --git a/docs/verbs.rst b/docs/verbs.rst
@@ -76,12 +76,14 @@ Returns a ``ResultDataset`` in a notebook context.
            $ hyrax test
 
 
-``umap``
---------
-Run `UMAP <https://umap-learn.readthedocs.io>`_ on the output of inference
-to reduce high-dimensional embeddings to 2D (or 3D) for visualization. By
-default, Hyrax uses the most recent inference output. See the
-:doc:`UMAP notebook </pre_executed/using_umap>` for configuration options.
+``reduce_dimensions``
+---------------------
+Run dimensionality-reduction algorithm `UMAP <https://umap-learn.readthedocs.io>`_ ,
+`t-SNE <https://scikit-learn.org/stable/modules/generated/sklearn.manifold.TSNE.html>`_,
+or `PCA <https://scikit-learn.org/stable/modules/generated/sklearn.decomposition.PCA.html>`_
+on the output of inference to reduce high-dimensional embeddings to 2D (or 3D) for visualization.
+By default, Hyrax uses the most recent inference output. See the
+:doc:`reduce_dimensions notebook </pre_executed/using_reduce_dimensions>` for configuration options.
 
 Returns a ``ResultDataset`` containing the reduced embeddings.
 
@@ -91,13 +93,13 @@ Returns a ``ResultDataset`` containing the reduced embeddings.
 
         .. code-block:: python
 
-           h.umap()
+           h.reduce_dimensions()
 
     .. tab-item:: CLI
 
         .. code-block:: bash
 
-           $ hyrax umap [-i <path_to_inference_output>]
+           $ hyrax reduce_dimensions [-i <path_to_inference_output>]
 
 
 ``visualize``

diff --git a/pyproject.toml b/pyproject.toml
@@ -45,7 +45,8 @@ dependencies = [
     "scipy", # Used in Holoviews visualization prototype
     "cython", # Used in Holoviews visualization prototype
     "mlflow >= 3.7", # Used to log training metrics and compare models
-    "umap-learn", # Used to map latent spaces down to 2d
+    "scikit-learn", # Used to map latent spaces down for t-sne and pca
+    "umap-learn", # Used to map latent spaces down to 2d with umap
     "pooch", # Used to download data files
     "pydantic>=2.13.4,<3.0", # Used for typed configuration schemas
     "onnx", # Used to export models to ONNX format