| blob | 5686dcf1f4747b00ef2a47d13aa9d7d3eb20708b |
1 {
2 "cells": [
3 {
4 "cell_type": "markdown",
5 "id": "244c2fb0-4339-476c-a2db-a641e124e25a",
6 "metadata": {},
7 "source": [
8 "# v2.1 exploration trying to make it work better"
9 ]
10 },
11 {
12 "cell_type": "code",
13 "execution_count": null,
14 "id": "e6cc7920-e380-4b81-bac0-cd6840450e9a",
15 "metadata": {},
16 "outputs": [],
17 "source": [
18 "# Environment\n",
19 "import os\n",
20 "import os.path as osp\n",
21 "import numpy as np\n",
22 "import pandas as pd\n",
23 "import tensorflow as tf\n",
24 "import matplotlib.pyplot as plt\n",
25 "import sys\n",
26 "# Local modules\n",
27 "sys.path.append('..')\n",
28 "import reproducibility\n",
29 "import pandas as pd\n",
30 "from utils import print_dict_summary\n",
31 "from data_funcs import rmse, process_train_dict\n",
32 "from moisture_rnn import RNNParams, RNNData, RNN, RNN_LSTM\n",
33 "from moisture_rnn_pkl import pkl2train\n",
34 "from tensorflow.keras.callbacks import Callback\n",
35 "from utils import hash2\n",
36 "import copy\n",
37 "import logging\n",
38 "import pickle\n",
39 "from utils import logging_setup, read_yml, read_pkl, hash_ndarray, hash_weights\n",
40 "import yaml\n",
41 "import copy"
42 ]
43 },
44 {
45 "cell_type": "code",
46 "execution_count": null,
47 "id": "f58e8839-bf0e-4995-b966-c09e4df001ce",
48 "metadata": {},
49 "outputs": [],
50 "source": [
51 "logging_setup()"
52 ]
53 },
54 {
55 "cell_type": "markdown",
56 "id": "fae67b50-f916-45a7-bcc7-61995ba39449",
57 "metadata": {},
58 "source": [
59 "## Tests"
60 ]
61 },
62 {
63 "cell_type": "markdown",
64 "id": "6322f0bc-107d-40a5-96dc-804495085a99",
65 "metadata": {
66 "jp-MarkdownHeadingCollapsed": true
67 },
68 "source": [
69 "## Test Other ML"
70 ]
71 },
72 {
73 "cell_type": "code",
74 "execution_count": null,
75 "id": "12992b9a-407f-4131-ac61-e1dc338386bf",
76 "metadata": {},
77 "outputs": [],
78 "source": [
79 "params = read_yml(\"params.yaml\", subkey='xgb')\n",
80 "params"
81 ]
82 },
83 {
84 "cell_type": "code",
85 "execution_count": null,
86 "id": "f214fdf8-bb76-4912-8f8c-5d0c8c1230c2",
87 "metadata": {},
88 "outputs": [],
89 "source": [
90 "dat = read_pkl(\"data/train.pkl\")"
91 ]
92 },
93 {
94 "cell_type": "code",
95 "execution_count": null,
96 "id": "888b7805-15f6-4c09-a05b-7aed7d253f6e",
97 "metadata": {},
98 "outputs": [],
99 "source": [
100 "cases = [*dat.keys()]"
101 ]
102 },
103 {
104 "cell_type": "code",
105 "execution_count": null,
106 "id": "375055d8-c070-4639-9561-e47d3f21f1f8",
107 "metadata": {},
108 "outputs": [],
109 "source": [
110 "rnn_dat = RNNData(dat[cases[10]], params['scaler'], params['features_list'])\n",
111 "rnn_dat.train_test_split(\n",
112 " time_fracs = [.8, .1, .1]\n",
113 ")\n",
114 "rnn_dat.scale_data()"
115 ]
116 },
117 {
118 "cell_type": "code",
119 "execution_count": null,
120 "id": "e79f8dc8-5cf8-4190-b4ff-e640f61bd78b",
121 "metadata": {},
122 "outputs": [],
123 "source": [
124 "from moisture_models import XGB, RF, LM"
125 ]
126 },
127 {
128 "cell_type": "code",
129 "execution_count": null,
130 "id": "b3aeb47f-261e-4e29-9eeb-67215e5628f6",
131 "metadata": {},
132 "outputs": [],
133 "source": [
134 "mod = XGB(params)"
135 ]
136 },
137 {
138 "cell_type": "code",
139 "execution_count": null,
140 "id": "cae9a20d-1caf-45aa-a9c4-aef21b65d9c8",
141 "metadata": {},
142 "outputs": [],
143 "source": [
144 "mod.params"
145 ]
146 },
147 {
148 "cell_type": "code",
149 "execution_count": null,
150 "id": "68a07b25-c586-4fc4-a3d5-c857354e7a2c",
151 "metadata": {},
152 "outputs": [],
153 "source": [
154 "mod.fit(rnn_dat.X_train, rnn_dat.y_train)"
155 ]
156 },
157 {
158 "cell_type": "code",
159 "execution_count": null,
160 "id": "c8f88819-0a7a-4420-abb9-56a47015a4de",
161 "metadata": {},
162 "outputs": [],
163 "source": [
164 "preds = mod.predict(rnn_dat.X_test)"
165 ]
166 },
167 {
168 "cell_type": "code",
169 "execution_count": null,
170 "id": "cb7cdf14-74d6-45e4-bc1b-7d4d47dd41ac",
171 "metadata": {},
172 "outputs": [],
173 "source": [
174 "rmse(preds, rnn_dat.y_test)"
175 ]
176 },
177 {
178 "cell_type": "code",
179 "execution_count": null,
180 "id": "74d478c7-8c01-448e-9a00-dd0e1ee8e325",
181 "metadata": {},
182 "outputs": [],
183 "source": [
184 "plt.plot(rnn_dat.y_test)\n",
185 "plt.plot(preds)"
186 ]
187 },
188 {
189 "cell_type": "code",
190 "execution_count": null,
191 "id": "c5441014-c39a-4414-a779-95b81e1ed6a8",
192 "metadata": {},
193 "outputs": [],
194 "source": [
195 "params = read_yml(\"params.yaml\", subkey='rf')\n",
196 "rnn_dat = RNNData(dat[cases[10]], features_list = ['Ed', 'Ew', 'solar', 'wind', 'rain'])\n",
197 "rnn_dat.train_test_split(\n",
198 " time_fracs = [.8, .1, .1]\n",
199 ")"
200 ]
201 },
202 {
203 "cell_type": "code",
204 "execution_count": null,
205 "id": "cafe711a-20cb-4bd3-a4bc-4995a843a021",
206 "metadata": {},
207 "outputs": [],
208 "source": [
209 "import importlib\n",
210 "import moisture_models\n",
211 "importlib.reload(moisture_models)"
212 ]
213 },
214 {
215 "cell_type": "code",
216 "execution_count": null,
217 "id": "ee45f7d6-f57f-4ff6-995a-527565565f94",
218 "metadata": {},
219 "outputs": [],
220 "source": [
221 "params"
222 ]
223 },
224 {
225 "cell_type": "code",
226 "execution_count": null,
227 "id": "fafe76e5-0212-4bd1-a058-535935a08780",
228 "metadata": {},
229 "outputs": [],
230 "source": [
231 "mod2 = RF(params)\n",
232 "mod2.fit(rnn_dat.X_train, rnn_dat.y_train.flatten())\n",
233 "preds2 = mod2.predict(rnn_dat.X_test)\n",
234 "print(rmse(preds2, rnn_dat.y_test.flatten()))\n",
235 "plt.plot(rnn_dat.y_test)\n",
236 "plt.plot(preds2)"
237 ]
238 },
239 {
240 "cell_type": "code",
241 "execution_count": null,
242 "id": "c0ab4244-996c-49af-bf4a-8b0c47b0b6db",
243 "metadata": {},
244 "outputs": [],
245 "source": [
246 "from moisture_models import RF\n",
247 "mod2 = RF(params)"
248 ]
249 },
250 {
251 "cell_type": "code",
252 "execution_count": null,
253 "id": "aa6c33fd-db35-4c77-9eee-fdb39a934959",
254 "metadata": {},
255 "outputs": [],
256 "source": []
257 },
258 {
259 "cell_type": "code",
260 "execution_count": null,
261 "id": "c5598bfe-2d87-4d23-869e-aff127782462",
262 "metadata": {},
263 "outputs": [],
264 "source": [
265 "params = read_yml(\"params.yaml\", subkey='lm')\n",
266 "rnn_dat = RNNData(dat[cases[10]], features_list = ['Ed', 'Ew', 'solar', 'wind', 'rain'])\n",
267 "rnn_dat.train_test_split(\n",
268 " time_fracs = [.8, .1, .1]\n",
269 ")\n",
270 "mod = LM(params)"
271 ]
272 },
273 {
274 "cell_type": "code",
275 "execution_count": null,
276 "id": "d828c15c-4078-4967-abff-c1fd15d4696d",
277 "metadata": {},
278 "outputs": [],
279 "source": [
280 "mod.fit(rnn_dat.X_train, rnn_dat.y_train)\n",
281 "preds = mod.predict(rnn_dat.X_test)\n",
282 "print(rmse(preds2, rnn_dat.y_test.flatten()))"
283 ]
284 },
285 {
286 "cell_type": "code",
287 "execution_count": null,
288 "id": "8496a32a-8269-4d6b-953e-7f33fe626789",
289 "metadata": {},
290 "outputs": [],
291 "source": []
292 },
293 {
294 "cell_type": "code",
295 "execution_count": null,
296 "id": "75ce8bf3-6efb-4dc7-b895-def92f6ce6b4",
297 "metadata": {},
298 "outputs": [],
299 "source": []
300 },
301 {
302 "cell_type": "markdown",
303 "id": "282cb651-b21f-401d-94c5-9e07530a9ba8",
304 "metadata": {},
305 "source": [
306 "## RNN"
307 ]
308 },
309 {
310 "cell_type": "code",
311 "execution_count": null,
312 "id": "8162841a-131b-4da6-a0c5-9d131a7cadf9",
313 "metadata": {},
314 "outputs": [],
315 "source": [
316 "dat = read_pkl(\"data/train.pkl\")"
317 ]
318 },
319 {
320 "cell_type": "code",
321 "execution_count": null,
322 "id": "96fe971b-c6d3-45ee-94ee-e4f426735d56",
323 "metadata": {},
324 "outputs": [],
325 "source": [
326 "params = RNNParams(read_yml(\"params.yaml\", subkey='rnn'))\n",
327 "params.update({\n",
328 " 'features_list': ['Ed', 'Ew', 'solar', 'wind', 'rain']\n",
329 "})"
330 ]
331 },
332 {
333 "cell_type": "code",
334 "execution_count": null,
335 "id": "5a55e8e7-1869-43fc-9bc6-09bd4f5a8d76",
336 "metadata": {},
337 "outputs": [],
338 "source": [
339 "rnn_dat2 = RNNData(dat[\"GSHPN_202401_set_2\"], params['scaler'], params['features_list'])\n",
340 "rnn_dat2.train_test_split(\n",
341 " time_fracs = [.8, .1, .1]\n",
342 ")\n",
343 "rnn_dat2.scale_data()\n",
344 "rnn_dat2.batch_reshape(timesteps = params['timesteps'], batch_size = params['batch_size'])"
345 ]
346 },
347 {
348 "cell_type": "code",
349 "execution_count": null,
350 "id": "aaec14ac-c6a6-4fcd-ad8e-d28143b92623",
351 "metadata": {},
352 "outputs": [],
353 "source": [
354 "reproducibility.set_seed()\n",
355 "rnn = RNN(params)\n",
356 "m, errs = rnn.run_model(rnn_dat2, plot_period=\"predict\")"
357 ]
358 },
359 {
360 "cell_type": "code",
361 "execution_count": null,
362 "id": "e5a1b91d-c692-484a-9e7c-c39cc5a4b1f7",
363 "metadata": {},
364 "outputs": [],
365 "source": []
366 },
367 {
368 "cell_type": "code",
369 "execution_count": null,
370 "id": "2a16aa0f-bd62-458f-94ba-d66fe1ec33cb",
371 "metadata": {},
372 "outputs": [],
373 "source": [
374 "params = RNNParams(read_yml(\"params.yaml\", subkey='rnn'))"
375 ]
376 },
377 {
378 "cell_type": "code",
379 "execution_count": null,
380 "id": "7ce63a46-47bd-47d6-87a0-573a97ca2880",
381 "metadata": {},
382 "outputs": [],
383 "source": [
384 "params = RNNParams(params)\n",
385 "params.update({'epochs': 200, \n",
386 " 'learning_rate': 0.001,\n",
387 " 'activation': ['tanh', 'tanh'], # Activation for RNN Layers, Dense layers respectively.\n",
388 " 'recurrent_layers': 2, 'recurrent_units': 30, \n",
389 " 'dense_layers': 2, 'dense_units': 30,\n",
390 " 'early_stopping_patience': 30, # how many epochs of no validation accuracy gain to wait before stopping\n",
391 " 'batch_schedule_type': 'exp', # Hidden state batch reset schedule\n",
392 " 'bmin': 20, # Lower bound of hidden state batch reset, \n",
393 " 'bmax': rnn_dat2.hours, # Upper bound of hidden state batch reset, using max hours\n",
394 " 'features_list': ['Ed', 'Ew', 'rain', 'solar', 'wind'],\n",
395 " 'timesteps': 12\n",
396 " })"
397 ]
398 },
399 {
400 "cell_type": "code",
401 "execution_count": null,
402 "id": "c2906b3f-6266-4bbc-9494-88ab36af58f9",
403 "metadata": {},
404 "outputs": [],
405 "source": [
406 "rnn_dat = RNNData(dat[\"FPRO3_202401_set_2\"], params['scaler'], params['features_list'])\n",
407 "rnn_dat.train_test_split(\n",
408 " time_fracs = [.8, .1, .1]\n",
409 ")\n",
410 "rnn_dat.scale_data()\n",
411 "rnn_dat.batch_reshape(timesteps = params['timesteps'], batch_size = params['batch_size'])"
412 ]
413 },
414 {
415 "cell_type": "code",
416 "execution_count": null,
417 "id": "c79ed028-ba60-4db5-9864-d3b2c01e09c3",
418 "metadata": {},
419 "outputs": [],
420 "source": [
421 "reproducibility.set_seed()\n",
422 "rnn = RNN(params)\n",
423 "m, errs = rnn.run_model(rnn_dat)"
424 ]
425 },
426 {
427 "cell_type": "code",
428 "execution_count": null,
429 "id": "947c1581-8021-48de-81fb-2212c1c17253",
430 "metadata": {},
431 "outputs": [],
432 "source": [
433 "rnn_dat = RNNData(dat[\"GSHPN_202401_set_2\"], params['scaler'], params['features_list'])\n",
434 "rnn_dat.train_test_split(\n",
435 " time_fracs = [.8, .1, .1]\n",
436 ")\n",
437 "rnn_dat.scale_data()\n",
438 "rnn_dat.batch_reshape(timesteps = params['timesteps'], batch_size = params['batch_size'])"
439 ]
440 },
441 {
442 "cell_type": "code",
443 "execution_count": null,
444 "id": "c134690e-599f-420e-b634-71a9201f0e83",
445 "metadata": {},
446 "outputs": [],
447 "source": [
448 "reproducibility.set_seed()\n",
449 "rnn = RNN(params)\n",
450 "m, errs = rnn.run_model(rnn_dat)"
451 ]
452 },
453 {
454 "cell_type": "code",
455 "execution_count": null,
456 "id": "cd6c8267-a98a-4fa8-9e42-54aebf15b0ee",
457 "metadata": {},
458 "outputs": [],
459 "source": []
460 },
461 {
462 "cell_type": "code",
463 "execution_count": null,
464 "id": "f76779c2-88db-4483-9191-b7902d2fb6a5",
465 "metadata": {},
466 "outputs": [],
467 "source": []
468 },
469 {
470 "cell_type": "markdown",
471 "id": "5ef092ff-8af1-491a-b0bf-cc3e674330e0",
472 "metadata": {},
473 "source": [
474 "## Phys Initialized"
475 ]
476 },
477 {
478 "cell_type": "code",
479 "execution_count": null,
480 "id": "5488628e-4552-4909-83e9-413fd6878bdd",
481 "metadata": {},
482 "outputs": [],
483 "source": [
484 "params.update({\n",
485 " 'epochs':100,\n",
486 " 'dense_layers': 0,\n",
487 " 'activation': ['relu', 'relu'],\n",
488 " 'phys_initialize': False,\n",
489 " 'dropout': [0,0]\n",
490 "})"
491 ]
492 },
493 {
494 "cell_type": "code",
495 "execution_count": null,
496 "id": "56bdf26c-07e7-4e4a-a567-af7dd0f564d9",
497 "metadata": {},
498 "outputs": [],
499 "source": [
500 "reproducibility.set_seed()\n",
501 "rnn = RNN(params)\n",
502 "m, errs = rnn.run_model(rnn_dat)"
503 ]
504 },
505 {
506 "cell_type": "code",
507 "execution_count": null,
508 "id": "01227b79-98f3-4931-bdfc-ff08afa8be5f",
509 "metadata": {},
510 "outputs": [],
511 "source": [
512 "rnn.model_train.summary()"
513 ]
514 },
515 {
516 "cell_type": "code",
517 "execution_count": null,
518 "id": "918a8bf0-638b-4b4b-82fe-c6a1965a72dd",
519 "metadata": {},
520 "outputs": [],
521 "source": []
522 },
523 {
524 "cell_type": "code",
525 "execution_count": null,
526 "id": "0aab34c7-8a09-480a-9d3e-619f7cf82b34",
527 "metadata": {},
528 "outputs": [],
529 "source": [
530 "params.update({\n",
531 " 'phys_initialize': True,\n",
532 " 'scaler': None, # TODO\n",
533 " 'dense_layers': 0, # NOT including single Dense output layer which is hard-coded\n",
534 " 'activation': ['linear', 'linear'], # TODO tanh, relu the same\n",
535 " 'batch_schedule_type': None # Hopefully this isn't a necessity like before, but maybe it will help\n",
536 "})"
537 ]
538 },
539 {
540 "cell_type": "code",
541 "execution_count": null,
542 "id": "ab549075-f71f-42ad-b36f-3d1e90247e33",
543 "metadata": {},
544 "outputs": [],
545 "source": [
546 "rnn_dat2 = RNNData(dat[cases[10]], params['scaler'], params['features_list'])\n",
547 "rnn_dat2.train_test_split(\n",
548 " time_fracs = [.8, .1, .1]\n",
549 ")\n",
550 "rnn_dat2.batch_reshape(timesteps = params['timesteps'], batch_size = params['batch_size'])"
551 ]
552 },
553 {
554 "cell_type": "code",
555 "execution_count": null,
556 "id": "195f337a-ac8a-4471-8226-94863b9385e2",
557 "metadata": {},
558 "outputs": [],
559 "source": [
560 "import importlib\n",
561 "import moisture_rnn\n",
562 "importlib.reload(moisture_rnn)\n",
563 "from moisture_rnn import RNN, RNNData"
564 ]
565 },
566 {
567 "cell_type": "code",
568 "execution_count": null,
569 "id": "9395d147-17a5-44ba-aaa2-a213ffde062b",
570 "metadata": {
571 "scrolled": true
572 },
573 "outputs": [],
574 "source": [
575 "reproducibility.set_seed()\n",
576 "\n",
577 "rnn = RNN(params)"
578 ]
579 },
580 {
581 "cell_type": "code",
582 "execution_count": null,
583 "id": "d3eebe8a-ff12-454b-81b6-6a138924f127",
584 "metadata": {},
585 "outputs": [],
586 "source": [
587 "m, errs = rnn.run_model(rnn_dat2)"
588 ]
589 },
590 {
591 "cell_type": "code",
592 "execution_count": null,
593 "id": "bcbb0159-74c5-4f56-9d69-d85a58ddbd1a",
594 "metadata": {},
595 "outputs": [],
596 "source": [
597 "rnn.model_predict.get_weights()"
598 ]
599 },
600 {
601 "cell_type": "code",
602 "execution_count": null,
603 "id": "c25f741a-6280-4cf2-8017-e56672236fdb",
604 "metadata": {},
605 "outputs": [],
606 "source": []
607 },
608 {
609 "cell_type": "code",
610 "execution_count": null,
611 "id": "e8ed2b03-6123-4bdf-9e26-ef2ce4951663",
612 "metadata": {},
613 "outputs": [],
614 "source": [
615 "params['rnn_units']"
616 ]
617 },
618 {
619 "cell_type": "code",
620 "execution_count": null,
621 "id": "e44302bf-af49-4140-ae31-54f7c88a6735",
622 "metadata": {},
623 "outputs": [],
624 "source": [
625 "params.update({\n",
626 " 'phys_initialize': True,\n",
627 " 'scaler': None, # TODO\n",
628 " 'dense_layers': 0, # NOT including single Dense output layer which is hard-coded\n",
629 " 'activation': ['relu', 'relu'], # TODO tanh, relu the same\n",
630 " 'batch_schedule_type': None # Hopefully this isn't a necessity like before, but maybe it will help\n",
631 "})"
632 ]
633 },
634 {
635 "cell_type": "code",
636 "execution_count": null,
637 "id": "9a8ac32d-551c-43e8-988e-a3b13e6d9cd9",
638 "metadata": {},
639 "outputs": [],
640 "source": [
641 "rnn_dat2 = RNNData(dat[cases[10]], params['scaler'], params['features_list'])\n",
642 "rnn_dat2.train_test_split(\n",
643 " time_fracs = [.8, .1, .1]\n",
644 ")\n",
645 "rnn_dat2.batch_reshape(timesteps = params['timesteps'], batch_size = params['batch_size'])"
646 ]
647 },
648 {
649 "cell_type": "code",
650 "execution_count": null,
651 "id": "ff727da8-38fb-4fda-999b-f712b98de0df",
652 "metadata": {
653 "scrolled": true
654 },
655 "outputs": [],
656 "source": [
657 "reproducibility.set_seed()\n",
658 "\n",
659 "rnn = RNN(params)\n",
660 "m, errs = rnn.run_model(rnn_dat2)"
661 ]
662 },
663 {
664 "cell_type": "code",
665 "execution_count": null,
666 "id": "b165074c-ea88-4b4d-8e41-6b6f22b4d221",
667 "metadata": {},
668 "outputs": [],
669 "source": []
670 },
671 {
672 "cell_type": "code",
673 "execution_count": null,
674 "id": "aa5cd4e6-4441-4c77-a086-e9edefbeb83b",
675 "metadata": {},
676 "outputs": [],
677 "source": []
678 },
679 {
680 "cell_type": "code",
681 "execution_count": null,
682 "id": "7bd1e05b-5cd8-48b4-8469-4842313d6097",
683 "metadata": {},
684 "outputs": [],
685 "source": []
686 },
687 {
688 "cell_type": "code",
689 "execution_count": null,
690 "id": "b399346d-20b8-4c97-898a-606a4be98065",
691 "metadata": {},
692 "outputs": [],
693 "source": []
694 },
695 {
696 "cell_type": "code",
697 "execution_count": null,
698 "id": "521285e6-6b6a-4d23-b688-9eb84b8eab68",
699 "metadata": {},
700 "outputs": [],
701 "source": []
702 },
703 {
704 "cell_type": "code",
705 "execution_count": null,
706 "id": "12c66af1-54fd-4398-8ee2-36eeb937c40d",
707 "metadata": {},
708 "outputs": [],
709 "source": []
710 },
711 {
712 "cell_type": "code",
713 "execution_count": null,
714 "id": "eb21fb8e-05c6-4a39-bdf1-4a57067c786d",
715 "metadata": {},
716 "outputs": [],
717 "source": []
718 },
719 {
720 "cell_type": "code",
721 "execution_count": null,
722 "id": "628a9105-ca06-44c4-ad00-13808e2f4773",
723 "metadata": {},
724 "outputs": [],
725 "source": []
726 },
727 {
728 "cell_type": "code",
729 "execution_count": null,
730 "id": "37fdbb3a-3e83-4541-93b2-982b6d4cbe93",
731 "metadata": {},
732 "outputs": [],
733 "source": []
734 },
735 {
736 "cell_type": "code",
737 "execution_count": null,
738 "id": "a592a4c9-cb3b-4174-8eaa-02afd00a1897",
739 "metadata": {},
740 "outputs": [],
741 "source": []
742 },
743 {
744 "cell_type": "code",
745 "execution_count": null,
746 "id": "3832fb05-417c-4648-8e2e-7748c06b3768",
747 "metadata": {},
748 "outputs": [],
749 "source": []
750 },
751 {
752 "cell_type": "markdown",
753 "id": "d2360aef-e9c4-4a71-922d-336e53b82537",
754 "metadata": {},
755 "source": [
756 "## LSTM\n",
757 "\n",
758 "TODO: FIX BELOW"
759 ]
760 },
761 {
762 "cell_type": "code",
763 "execution_count": null,
764 "id": "71d4e441-9bf1-4d57-bb37-091553e23212",
765 "metadata": {},
766 "outputs": [],
767 "source": [
768 "import importlib \n",
769 "import moisture_rnn\n",
770 "importlib.reload(moisture_rnn)\n",
771 "from moisture_rnn import RNN_LSTM"
772 ]
773 },
774 {
775 "cell_type": "code",
776 "execution_count": null,
777 "id": "0f6ba896-e3be-4a9f-8a42-3df64aff7d63",
778 "metadata": {},
779 "outputs": [],
780 "source": [
781 "params = read_yml(\"params.yaml\", subkey=\"lstm\")\n",
782 "params = RNNParams(params)"
783 ]
784 },
785 {
786 "cell_type": "code",
787 "execution_count": null,
788 "id": "a4cf567e-d623-4e14-b578-eed88b80d04e",
789 "metadata": {},
790 "outputs": [],
791 "source": [
792 "rnn_dat = RNNData(dat[cases[10]], params['scaler'], params['features_list'])\n",
793 "rnn_dat.train_test_split(\n",
794 " time_fracs = [.8, .1, .1]\n",
795 ")\n",
796 "rnn_dat.scale_data()\n",
797 "rnn_dat.batch_reshape(timesteps = params['timesteps'], batch_size = params['batch_size'])"
798 ]
799 },
800 {
801 "cell_type": "code",
802 "execution_count": null,
803 "id": "0157a6bc-3a99-4b87-a42c-ab770d19ae37",
804 "metadata": {},
805 "outputs": [],
806 "source": [
807 "from moisture_rnn import ResetStatesCallback, EarlyStoppingCallback\n",
808 "params.update({'epochs': 50, 'learning_rate': 0.001, 'verbose_fit': True, 'rnn_layers': 2, 'rnn_units': 20, 'dense_layers': 1, 'dense_units': 10,\n",
809 " 'activation': ['tanh', 'tanh'], 'features_list': rnn_dat.features_list,\n",
810 " 'batch_schedule_type':'exp', 'bmin': 10, 'bmax':rnn_dat.hours})\n",
811 "reproducibility.set_seed(123)\n",
812 "lstm = RNN_LSTM(params)\n",
813 "\n",
814 "history = lstm.model_train.fit(rnn_dat.X_train, rnn_dat.y_train, \n",
815 " batch_size = params['batch_size'], epochs=params['epochs'], \n",
816 " callbacks = [ResetStatesCallback(params),\n",
817 " EarlyStoppingCallback(patience = 15)],\n",
818 " validation_data = (rnn_dat.X_val, rnn_dat.y_val))\n",
819 " "
820 ]
821 },
822 {
823 "cell_type": "code",
824 "execution_count": null,
825 "id": "ec95e7d4-6d57-441b-b673-f10625ee5dec",
826 "metadata": {},
827 "outputs": [],
828 "source": []
829 },
830 {
831 "cell_type": "code",
832 "execution_count": null,
833 "id": "9b3c8d8d-ea50-44ea-8c0c-414e07cd01ac",
834 "metadata": {},
835 "outputs": [],
836 "source": []
837 },
838 {
839 "cell_type": "code",
840 "execution_count": null,
841 "id": "03063e3c-e8f4-451d-b0cf-25bd965cd9d6",
842 "metadata": {},
843 "outputs": [],
844 "source": [
845 "params.update({'epochs': 50, 'learning_rate': 0.001, 'verbose_fit': True, 'rnn_layers': 2, 'rnn_units': 20, 'dense_layers': 1, 'dense_units': 10,\n",
846 " 'activation': ['tanh', 'tanh'], 'features_list': rnn_dat.features_list,\n",
847 " 'batch_schedule_type':'exp', 'bmin': 10, 'bmax':rnn_dat.hours,\n",
848 " 'early_stopping_patience': 25})\n",
849 "reproducibility.set_seed(123)\n",
850 "lstm = RNN_LSTM(params)\n",
851 "m, errs = lstm.run_model(rnn_dat)"
852 ]
853 },
854 {
855 "cell_type": "code",
856 "execution_count": null,
857 "id": "f60a24c6-9a67-45aa-bc5c-8818aa0ca049",
858 "metadata": {},
859 "outputs": [],
860 "source": []
861 },
862 {
863 "cell_type": "code",
864 "execution_count": null,
865 "id": "00910bd2-f050-438c-ab3b-c793b83cb5f5",
866 "metadata": {},
867 "outputs": [],
868 "source": [
869 "rnn_dat.spatial"
870 ]
871 },
872 {
873 "cell_type": "code",
874 "execution_count": null,
875 "id": "236b33e3-e864-4453-be16-cf07338c4105",
876 "metadata": {},
877 "outputs": [],
878 "source": [
879 "params = RNNParams(read_yml(\"params.yaml\", subkey='lstm'))\n",
880 "params"
881 ]
882 },
883 {
884 "cell_type": "code",
885 "execution_count": null,
886 "id": "fe2a484c-dc99-45a9-89fc-2f451bd719b5",
887 "metadata": {},
888 "outputs": [],
889 "source": [
890 "train = read_pkl(\"data/train.pkl\")"
891 ]
892 },
893 {
894 "cell_type": "code",
895 "execution_count": null,
896 "id": "07bfac87-a6d4-4dcc-8d11-adf83eafab76",
897 "metadata": {},
898 "outputs": [],
899 "source": [
900 "from itertools import islice\n",
901 "train = {k: train[k] for k in islice(train, 100)}"
902 ]
903 },
904 {
905 "cell_type": "code",
906 "execution_count": null,
907 "id": "4e26099b-f760-4047-afec-9e751d24b7a6",
908 "metadata": {},
909 "outputs": [],
910 "source": [
911 "from data_funcs import combine_nested\n",
912 "rnn_dat_sp = RNNData(\n",
913 " combine_nested(train), # input dictionary\n",
914 " scaler=\"standard\", # data scaling type\n",
915 " features_list = params['features_list'] # features for predicting outcome\n",
916 ")\n",
917 "\n",
918 "\n",
919 "rnn_dat_sp.train_test_split( \n",
920 " time_fracs = [.8, .1, .1], # Percent of total time steps used for train/val/test\n",
921 " space_fracs = [.8, .1, .1] # Percent of total timeseries used for train/val/test\n",
922 ")\n",
923 "rnn_dat_sp.scale_data()\n",
924 "\n",
925 "rnn_dat_sp.batch_reshape(\n",
926 " timesteps = params['timesteps'], # Timesteps aka sequence length for RNN input data. \n",
927 " batch_size = params['batch_size'] # Number of samples of length timesteps for a single round of grad. descent\n",
928 ")"
929 ]
930 },
931 {
932 "cell_type": "code",
933 "execution_count": null,
934 "id": "10738795-c83b-4da3-88ba-09278caa35f8",
935 "metadata": {},
936 "outputs": [],
937 "source": [
938 "params.update({\n",
939 " 'loc_batch_reset': rnn_dat_sp.n_seqs # Used to reset hidden state when location changes for a given batch\n",
940 "})"
941 ]
942 },
943 {
944 "cell_type": "code",
945 "execution_count": null,
946 "id": "9c5d45cc-bcf0-4b6c-9c51-c4c790a2d9a5",
947 "metadata": {},
948 "outputs": [],
949 "source": [
950 "rnn_sp = RNN_LSTM(params)\n",
951 "m_sp, errs = rnn_sp.run_model(rnn_dat_sp)"
952 ]
953 },
954 {
955 "cell_type": "code",
956 "execution_count": null,
957 "id": "ee332ccf-4e4a-4f66-b4d6-c079dbdb1411",
958 "metadata": {},
959 "outputs": [],
960 "source": [
961 "errs.mean()"
962 ]
963 },
964 {
965 "cell_type": "code",
966 "execution_count": null,
967 "id": "739d4b26-641e-47b2-a90a-67cd32215d05",
968 "metadata": {},
969 "outputs": [],
970 "source": []
971 }
972 ],
973 "metadata": {
974 "kernelspec": {
975 "display_name": "Python 3 (ipykernel)",
976 "language": "python",
977 "name": "python3"
978 },
979 "language_info": {
980 "codemirror_mode": {
981 "name": "ipython",
982 "version": 3
983 },
984 "file_extension": ".py",
985 "mimetype": "text/x-python",
986 "name": "python",
987 "nbconvert_exporter": "python",
988 "pygments_lexer": "ipython3",
989 "version": "3.12.6"
990 }
991 },
992 "nbformat": 4,
993 "nbformat_minor": 5
994 }