From 1a6ebb4c83306ebcfbfd0b7c15aacd4073048d2e Mon Sep 17 00:00:00 2001
From: jh-206 <jhirschi206@gmail.com>
Date: Tue, 22 Oct 2024 16:38:33 -0600
Subject: [PATCH] Update fmda_rnn_train_and_save.ipynb

---
 fmda/test_notebooks/fmda_rnn_train_and_save.ipynb | 25 ++++++++++++++++++-----
 1 file changed, 20 insertions(+), 5 deletions(-)

diff --git a/fmda/test_notebooks/fmda_rnn_train_and_save.ipynb b/fmda/test_notebooks/fmda_rnn_train_and_save.ipynb
index 354da31..8b396b0 100644
--- a/fmda/test_notebooks/fmda_rnn_train_and_save.ipynb
+++ b/fmda/test_notebooks/fmda_rnn_train_and_save.ipynb
@@ -5,9 +5,9 @@
    "id": "83b774b3-ef55-480a-b999-506676e49145",
    "metadata": {},
    "source": [
-    "# v2.2 run RNN with Spatial Training\n",
+    "# v2.3 run RNN and Save\n",
     "\n",
-    "This notebook is intended to set up a test where the RNN is run serial by location and compared to the spatial training scheme. Additionally, the ODE model with the augmented KF will be run as a comparison, but note that the RNN models will be predicting entirely without knowledge of the heldout locations, while the augmented KF will be run directly on the test locations.\n"
+    "This notebook is intended to test traing and then saving a model object for later use.\n"
    ]
   },
   {
@@ -114,7 +114,11 @@
    "source": [
     "# Params used for setting up RNN\n",
     "params = read_yml(\"../params.yaml\", subkey='rnn') \n",
-    "params"
+    "params.update({\n",
+    "    'hidden_layers': ['dense', 'lstm', 'attention', 'dense'],\n",
+    "    'hidden_units': [64, 32, None, 32],\n",
+    "    'hidden_activation': ['relu', 'tanh', None, 'relu']\n",
+    "})"
    ]
   },
   {
@@ -262,7 +266,7 @@
    "id": "62c1b049-304e-4c90-b1d2-b9b96b9a202f",
    "metadata": {},
    "source": [
-    "## Save Model"
+    "## Save "
    ]
   },
   {
@@ -274,7 +278,18 @@
    "source": [
     "outpath = \"../outputs/models\"\n",
     "filename = osp.join(outpath, f\"model_predict_raws_rocky.keras\")\n",
-    "rnn_sp.model_predict.save(filename)"
+    "rnn_sp.model_predict.save(filename) # save prediction model only"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "cf6231a9-0c7b-45ba-ac75-7fb5b6124c72",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "with open(f\"{outpath}/rnn_data_rocky.pkl\", 'wb') as file:\n",
+    "    pickle.dump(rnn_dat_sp, file)"
    ]
   },
   {
-- 
2.11.4.GIT