fmda/fmda_rnn_spatial.ipynb

   1 {
   2  "cells": [
   3   {
   4    "cell_type": "markdown",
   5    "id": "83b774b3-ef55-480a-b999-506676e49145",
   6    "metadata": {},
   7    "source": [
   8     "# v2.1 run RNN with Spatial Training\n",
   9     "\n",
  10     "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"
  11    ]
  12   },
  13   {
  14    "cell_type": "markdown",
  15    "id": "bbd84d61-a9cd-47b4-b538-4986fb10b98d",
  16    "metadata": {},
  17    "source": [
  18     "## Environment Setup"
  19    ]
  20   },
  21   {
  22    "cell_type": "code",
  23    "execution_count": null,
  24    "id": "83cc1dc4-3dcb-4325-9263-58101a3dc378",
  25    "metadata": {},
  26    "outputs": [],
  27    "source": [
  28     "import numpy as np\n",
  29     "from utils import print_dict_summary, print_first, str2time, logging_setup\n",
  30     "import pickle\n",
  31     "import logging\n",
  32     "import os.path as osp\n",
  33     "from moisture_rnn_pkl import pkl2train\n",
  34     "from moisture_rnn import RNNParams, RNNData, RNN \n",
  35     "from utils import hash2, read_yml, read_pkl, retrieve_url, Dict\n",
  36     "from moisture_rnn import RNN\n",
  37     "import reproducibility\n",
  38     "from data_funcs import rmse, to_json, combine_nested, build_train_dict\n",
  39     "from moisture_models import run_augmented_kf\n",
  40     "import copy\n",
  41     "import pandas as pd\n",
  42     "import matplotlib.pyplot as plt\n",
  43     "import yaml\n",
  44     "import time"
  45    ]
  46   },
  47   {
  48    "cell_type": "code",
  49    "execution_count": null,
  50    "id": "17db9b90-a931-4674-a447-5b8ffbcdc86a",
  51    "metadata": {},
  52    "outputs": [],
  53    "source": [
  54     "logging_setup()"
  55    ]
  56   },
  57   {
  58    "cell_type": "code",
  59    "execution_count": null,
  60    "id": "35319c1c-7849-4b8c-8262-f5aa6656e0c7",
  61    "metadata": {},
  62    "outputs": [],
  63    "source": [
  64     "retrieve_url(\n",
  65     "    url = \"https://demo.openwfm.org/web/data/fmda/dicts/fmda_nw_202401-05_f05.pkl\", \n",
  66     "    dest_path = \"data/fmda_nw_202401-05_f05.pkl\")"
  67    ]
  68   },
  69   {
  70    "cell_type": "code",
  71    "execution_count": null,
  72    "id": "eabdbd9c-07d9-4bae-9851-cca79f321895",
  73    "metadata": {},
  74    "outputs": [],
  75    "source": [
  76     "file_paths = ['data/fmda_nw_202401-05_f05.pkl']"
  77    ]
  78   },
  79   {
  80    "cell_type": "code",
  81    "execution_count": null,
  82    "id": "dcca6185-e799-4dd1-8acb-87ad33c411d7",
  83    "metadata": {},
  84    "outputs": [],
  85    "source": [
  86     "# read/write control\n",
  87     "train_file='data/train.pkl'\n",
  88     "train_create=True   # if false, read\n",
  89     "train_write=False\n",
  90     "train_read=False"
  91    ]
  92   },
  93   {
  94    "cell_type": "code",
  95    "execution_count": null,
  96    "id": "604388de-11ab-45c3-9f0d-80bdff0cca60",
  97    "metadata": {},
  98    "outputs": [],
  99    "source": [
 100     "# Params used for data filtering\n",
 101     "params_data = read_yml(\"params_data.yaml\") \n",
 102     "params_data"
 103    ]
 104   },
 105   {
 106    "cell_type": "code",
 107    "execution_count": null,
 108    "id": "211a1c2f-ba8d-40b8-b29c-daa38af97a26",
 109    "metadata": {},
 110    "outputs": [],
 111    "source": [
 112     "# Params used for setting up RNN\n",
 113     "params = read_yml(\"params.yaml\", subkey='rnn') \n",
 114     "params"
 115    ]
 116   },
 117   {
 118    "cell_type": "code",
 119    "execution_count": null,
 120    "id": "be81d76c-3123-4467-982b-d2da5b1c08bd",
 121    "metadata": {
 122     "scrolled": true
 123    },
 124    "outputs": [],
 125    "source": [
 126     "train = build_train_dict(file_paths, atm_source=\"HRRR\", params_data = params_data, spatial=False, verbose=True)"
 127    ]
 128   },
 129   {
 130    "cell_type": "code",
 131    "execution_count": null,
 132    "id": "bc0a775b-b587-42ef-8576-e36dc0be3a75",
 133    "metadata": {
 134     "scrolled": true
 135    },
 136    "outputs": [],
 137    "source": [
 138     "# if train_create:\n",
 139     "#     logging.info('creating the training cases from files %s',file_paths)\n",
 140     "#     # osp.join works on windows too, joins paths using \\ or /\n",
 141     "#     train = process_train_dict(file_paths, atm_dict=\"HRRR\", params_data = params_data, verbose=True)\n",
 142     "# if train_write:\n",
 143     "#     with open(train_file, 'wb') as file:\n",
 144     "#         logging.info('Writing the rain cases into file %s',train_file)\n",
 145     "#         pickle.dump(train, file)\n",
 146     "# if train_read:\n",
 147     "#     logging.info('Reading the train cases from file %s',train_file)\n",
 148     "#     train = read_pkl(train_file)"
 149    ]
 150   },
 151   {
 152    "cell_type": "code",
 153    "execution_count": null,
 154    "id": "23cd60c0-9865-4314-9a96-948c3d400c08",
 155    "metadata": {},
 156    "outputs": [],
 157    "source": [
 158     "from itertools import islice\n",
 159     "train = {k: train[k] for k in islice(train, 250)}"
 160    ]
 161   },
 162   {
 163    "cell_type": "markdown",
 164    "id": "efc10cdc-f18b-4781-84da-b8e2eef39981",
 165    "metadata": {},
 166    "source": [
 167     "## Setup Validation Runs"
 168    ]
 169   },
 170   {
 171    "cell_type": "markdown",
 172    "id": "2d9cd5c5-87ed-41f9-b36c-e0c18d58c841",
 173    "metadata": {},
 174    "source": [
 175     "The following parameters will be used for both serial and spatial models."
 176    ]
 177   },
 178   {
 179    "cell_type": "code",
 180    "execution_count": null,
 181    "id": "66f40c9f-c1c2-4b12-bf14-2ada8c26113d",
 182    "metadata": {},
 183    "outputs": [],
 184    "source": [
 185     "params = RNNParams(params)\n",
 186     "params.update({'epochs': 200, \n",
 187     "               'learning_rate': 0.001,\n",
 188     "               'activation': ['tanh', 'tanh'], # Activation for RNN Layers, Dense layers respectively.\n",
 189     "               'rnn_layers': 2, 'recurrent_units': 30, \n",
 190     "               'dense_layers': 2, 'dense_units': 30,\n",
 191     "               'early_stopping_patience': 30, # how many epochs of no validation accuracy gain to wait before stopping\n",
 192     "               'batch_schedule_type': 'exp', # Hidden state batch reset schedule\n",
 193     "               'bmin': 20, # Lower bound of hidden state batch reset, \n",
 194     "               'bmax': params_data['hours'], # Upper bound of hidden state batch reset, using max hours\n",
 195     "               'features_list': ['Ed', 'Ew', 'rain', 'elev', 'lon', 'lat', 'solar', 'wind'],\n",
 196     "               'timesteps': 12\n",
 197     "              })"
 198    ]
 199   },
 200   {
 201    "cell_type": "code",
 202    "execution_count": null,
 203    "id": "36823193-b93c-421e-b699-8c1ae5719309",
 204    "metadata": {},
 205    "outputs": [],
 206    "source": [
 207     "reproducibility.set_seed(123)"
 208    ]
 209   },
 210   {
 211    "cell_type": "markdown",
 212    "id": "a24d76fc-6c25-43e7-99df-3cd5dbf84fc3",
 213    "metadata": {},
 214    "source": [
 215     "## Spatial Data Training\n",
 216     "\n",
 217     "This method combines the training timeseries data into a single 3-d array, with timeseries at the same location arranged appropriately in the right order for a given `batch_size` hyperparameter. The hidden state of the recurrent layers are set up reset when the location changes. "
 218    ]
 219   },
 220   {
 221    "cell_type": "code",
 222    "execution_count": null,
 223    "id": "3b5371a9-c1e8-4df5-b360-210746f7cd52",
 224    "metadata": {},
 225    "outputs": [],
 226    "source": [
 227     "# Start timer for code \n",
 228     "start_time = time.time()"
 229    ]
 230   },
 231   {
 232    "cell_type": "code",
 233    "execution_count": null,
 234    "id": "faf93470-b55f-4770-9fa9-3288a2f13fcc",
 235    "metadata": {},
 236    "outputs": [],
 237    "source": [
 238     "# Combine Nested Dictionary into Spatial Data\n",
 239     "train_sp = Dict(combine_nested(train))"
 240    ]
 241   },
 242   {
 243    "cell_type": "code",
 244    "execution_count": null,
 245    "id": "c0c7f5fb-4c33-45f8-9a2e-38c9ab1cd4e3",
 246    "metadata": {},
 247    "outputs": [],
 248    "source": [
 249     "rnn_dat_sp = RNNData(\n",
 250     "    train_sp, # input dictionary\n",
 251     "    scaler=\"standard\",  # data scaling type\n",
 252     "    features_list = params['features_list'] # features for predicting outcome\n",
 253     ")\n",
 254     "\n",
 255     "\n",
 256     "rnn_dat_sp.train_test_split(   \n",
 257     "    time_fracs = [.8, .1, .1], # Percent of total time steps used for train/val/test\n",
 258     "    space_fracs = [.8, .1, .1] # Percent of total timeseries used for train/val/test\n",
 259     ")\n",
 260     "rnn_dat_sp.scale_data()\n",
 261     "\n",
 262     "rnn_dat_sp.batch_reshape(\n",
 263     "    timesteps = params['timesteps'], # Timesteps aka sequence length for RNN input data. \n",
 264     "    batch_size = params['batch_size'] # Number of samples of length timesteps for a single round of grad. descent\n",
 265     ")"
 266    ]
 267   },
 268   {
 269    "cell_type": "code",
 270    "execution_count": null,
 271    "id": "7431bc95-d384-40fd-a622-bbc0ee68e5cd",
 272    "metadata": {},
 273    "outputs": [],
 274    "source": [
 275     "# Update Params specific to spatial training\n",
 276     "params.update({\n",
 277     "    'loc_batch_reset': rnn_dat_sp.n_seqs # Used to reset hidden state when location changes for a given batch\n",
 278     "})"
 279    ]
 280   },
 281   {
 282    "cell_type": "code",
 283    "execution_count": null,
 284    "id": "4bc11474-fed8-47f2-b9cf-dfdda0d3d3b2",
 285    "metadata": {},
 286    "outputs": [],
 287    "source": [
 288     "rnn_sp = RNN(params)\n",
 289     "m_sp, errs = rnn_sp.run_model(rnn_dat_sp)"
 290    ]
 291   },
 292   {
 293    "cell_type": "code",
 294    "execution_count": null,
 295    "id": "704ad662-d81a-488d-be3d-e90bf775a5b8",
 296    "metadata": {},
 297    "outputs": [],
 298    "source": [
 299     "errs.mean()"
 300    ]
 301   },
 302   {
 303    "cell_type": "code",
 304    "execution_count": null,
 305    "id": "d53571e3-b6cf-49aa-9848-e3c77053283d",
 306    "metadata": {},
 307    "outputs": [],
 308    "source": [
 309     "# End Timer\n",
 310     "end_time = time.time()\n",
 311     "\n",
 312     "# Calculate Code Runtime\n",
 313     "elapsed_time_sp = end_time - start_time\n",
 314     "print(f\"Spatial Training Elapsed time: {elapsed_time_sp:.4f} seconds\")"
 315    ]
 316   },
 317   {
 318    "cell_type": "markdown",
 319    "id": "7d8292a2-418c-48ed-aff7-ccbe98b046d3",
 320    "metadata": {},
 321    "source": [
 322     "## Run ODE + KF "
 323    ]
 324   },
 325   {
 326    "cell_type": "code",
 327    "execution_count": null,
 328    "id": "cca12d8c-c0e1-4df4-b2ca-20440485f2f3",
 329    "metadata": {},
 330    "outputs": [],
 331    "source": [
 332     "# Get timeseries IDs from previous RNNData object\n",
 333     "test_cases = rnn_dat_sp.loc['test_locs']\n",
 334     "print(len(test_cases))"
 335    ]
 336   },
 337   {
 338    "cell_type": "code",
 339    "execution_count": null,
 340    "id": "997f2534-7e77-45b3-93bf-d988837dfc0b",
 341    "metadata": {},
 342    "outputs": [],
 343    "source": [
 344     "test_ind = rnn_dat_sp.test_ind # Time index for test period start\n",
 345     "print(test_ind)"
 346    ]
 347   },
 348   {
 349    "cell_type": "code",
 350    "execution_count": null,
 351    "id": "1e4ffc68-c775-41c6-ac42-f49c76824b43",
 352    "metadata": {
 353     "scrolled": true
 354    },
 355    "outputs": [],
 356    "source": [
 357     "outputs_kf = {}\n",
 358     "for case in test_cases:\n",
 359     "    print(\"~\"*50)\n",
 360     "    print(case)\n",
 361     "    # Run Augmented KF\n",
 362     "    print('Running Augmented KF')\n",
 363     "    train[case]['h2'] = test_ind\n",
 364     "    train[case]['scale_fm'] = 1\n",
 365     "    m, Ec = run_augmented_kf(train[case])\n",
 366     "    y = train[case]['y']        \n",
 367     "    train[case]['m_kf'] = m\n",
 368     "    print(f\"KF RMSE: {rmse(m[test_ind:],y[test_ind:])}\")\n",
 369     "    outputs_kf[case] = {'case':case, 'errs': rmse(m[test_ind:],y[test_ind:])}"
 370    ]
 371   },
 372   {
 373    "cell_type": "code",
 374    "execution_count": null,
 375    "id": "57b19ec5-23f6-44ec-9f71-16d4d69aec68",
 376    "metadata": {},
 377    "outputs": [],
 378    "source": [
 379     "df_kf = pd.DataFrame.from_dict(outputs_kf).transpose()\n",
 380     "df_kf.head()"
 381    ]
 382   },
 383   {
 384    "cell_type": "code",
 385    "execution_count": null,
 386    "id": "25a9d2fe-83f7-4ef3-a04b-14c970b6e2ba",
 387    "metadata": {},
 388    "outputs": [],
 389    "source": [
 390     "df_kf.errs.mean()"
 391    ]
 392   },
 393   {
 394    "cell_type": "markdown",
 395    "id": "86795281-f8ea-4141-81ea-c53fae830e80",
 396    "metadata": {},
 397    "source": [
 398     "## Compare"
 399    ]
 400   },
 401   {
 402    "cell_type": "code",
 403    "execution_count": null,
 404    "id": "508a6392-49bc-4471-ad8e-814f60119283",
 405    "metadata": {},
 406    "outputs": [],
 407    "source": [
 408     "print(f\"Total Test Cases: {len(test_cases)}\")\n",
 409     "print(f\"Total Test Hours: {rnn_dat_temp.y_test.shape[0]}\")"
 410    ]
 411   },
 412   {
 413    "cell_type": "code",
 414    "execution_count": null,
 415    "id": "73e8ca05-d17b-4e72-8def-fa77664e7bb0",
 416    "metadata": {},
 417    "outputs": [],
 418    "source": [
 419     "print(f\"Spatial Training RMSE: {errs.mean()}\")\n",
 420     "print(f\"Augmented KF RMSE: {df_kf.errs.mean()}\")"
 421    ]
 422   },
 423   {
 424    "cell_type": "code",
 425    "execution_count": null,
 426    "id": "a73d22ee-707b-44a3-80ab-ad6e671731cf",
 427    "metadata": {},
 428    "outputs": [],
 429    "source": []
 430   },
 431   {
 432    "cell_type": "code",
 433    "execution_count": null,
 434    "id": "272bfb32-e8e2-49dd-8f90-4b5b09c3a2a2",
 435    "metadata": {},
 436    "outputs": [],
 437    "source": [
 438     "print(f\"Spatial Training Elapsed time: {elapsed_time_sp:.4f} seconds\")"
 439    ]
 440   },
 441   {
 442    "cell_type": "code",
 443    "execution_count": null,
 444    "id": "38ab08fb-ac97-45be-8907-6f9cd124243b",
 445    "metadata": {},
 446    "outputs": [],
 447    "source": []
 448   }
 449  ],
 450  "metadata": {
 451   "kernelspec": {
 452    "display_name": "Python 3 (ipykernel)",
 453    "language": "python",
 454    "name": "python3"
 455   },
 456   "language_info": {
 457    "codemirror_mode": {
 458     "name": "ipython",
 459     "version": 3
 460    },
 461    "file_extension": ".py",
 462    "mimetype": "text/x-python",
 463    "name": "python",
 464    "nbconvert_exporter": "python",
 465    "pygments_lexer": "ipython3",
 466    "version": "3.12.5"
 467   }
 468  },
 469  "nbformat": 4,
 470  "nbformat_minor": 5
 471 }