File size: 24,304 Bytes
9a67fbe |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 |
{
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"id": "RqYoTBJrCcJx"
},
"outputs": [],
"source": [
"%%capture\n",
"!pip install torch_geometric rdkit\n",
"!pip install torch-scatter -f https://data.pyg.org/whl/torch-$(python -c \"import torch; print(torch.__version__)\").html"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"id": "M0suAYOT5R-N"
},
"outputs": [],
"source": [
"# Standard library\n",
"from collections import defaultdict\n",
"from typing import List, Tuple\n",
"import importlib.resources as pkg_resources\n",
"from multiprocessing.pool import ThreadPool\n",
"\n",
"# Third-party scientific stack\n",
"import numpy as np\n",
"import pandas as pd\n",
"from scipy.sparse import coo_matrix\n",
"from scipy.sparse.linalg import eigsh\n",
"\n",
"# RDKit\n",
"from rdkit import Chem\n",
"\n",
"# PyTorch core\n",
"import torch\n",
"import torch.nn as nn\n",
"import torch.nn.functional as F\n",
"\n",
"# PyTorch Geometric\n",
"import torch_geometric\n",
"from torch_geometric.data import Data, Batch, DataLoader\n",
"from torch_geometric.nn import GCNConv, GINConv, BatchNorm, global_mean_pool\n",
"from torch_geometric.loader import DataLoader as PyGDataLoader"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "4Wginy3UChb_",
"outputId": "981c9246-0f1c-403b-bac3-fc4941611c14"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
]
}
],
"source": [
"from google.colab import drive\n",
"drive.mount('/content/drive')"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"id": "obXQxOveCM9l"
},
"outputs": [],
"source": [
"def set_seed(seed):\n",
" \"\"\"\n",
" Fix all random seeds for reproducibility across Python, NumPy, and PyTorch.\n",
" \"\"\"\n",
" import random\n",
" random.seed(seed)\n",
" np.random.seed(seed)\n",
" torch.manual_seed(seed)\n",
" if torch.cuda.is_available():\n",
" torch.cuda.manual_seed_all(seed)\n",
" torch.backends.cudnn.deterministic = True\n",
" torch.backends.cudnn.benchmark = False\n",
"\n",
"set_seed(42)"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {
"id": "fYyEoAXuCM9m"
},
"outputs": [],
"source": [
"import torch\n",
"import torch.nn as nn\n",
"import torch.nn.functional as F\n",
"from torch_geometric.nn import (\n",
" PNAConv,\n",
" global_mean_pool,\n",
" global_max_pool,\n",
" global_add_pool\n",
")\n",
"from torch_geometric.utils import degree\n",
"\n",
"class PolyatomicNet(nn.Module):\n",
" def __init__(self, node_feat_dim, edge_feat_dim, graph_feat_dim,deg,\n",
" hidden_dim=128, num_layers=5, dropout=0.1):\n",
" super().__init__()\n",
"\n",
" self.node_emb = nn.Linear(node_feat_dim, hidden_dim)\n",
" self.deg = deg\n",
" self.virtualnode_emb = nn.Embedding(1, hidden_dim)\n",
" self.vn_mlp = nn.Sequential(\n",
" nn.Linear(hidden_dim, hidden_dim),\n",
" nn.ReLU(),\n",
" nn.Linear(hidden_dim, hidden_dim)\n",
" )\n",
"\n",
" # For graph-level feature projection\n",
" self.graph_proj = nn.Sequential(\n",
" nn.Linear(graph_feat_dim, hidden_dim),\n",
" nn.ReLU(),\n",
" nn.Linear(hidden_dim, hidden_dim)\n",
" )\n",
"\n",
" # PNAConv requires degree preprocessing\n",
" self.deg_emb = nn.Embedding(20, hidden_dim) # cap degree buckets\n",
"\n",
" aggregators = ['mean', 'min', 'max', 'std']\n",
" scalers = ['identity', 'amplification', 'attenuation']\n",
"\n",
" self.convs = nn.ModuleList()\n",
" self.bns = nn.ModuleList()\n",
"\n",
" for _ in range(num_layers):\n",
" conv = PNAConv(\n",
" in_channels=hidden_dim,\n",
" out_channels=hidden_dim,\n",
" aggregators=aggregators,\n",
" scalers=scalers,\n",
" edge_dim=edge_feat_dim,\n",
" towers=4,\n",
" pre_layers=1,\n",
" post_layers=1,\n",
" divide_input=True,\n",
" deg=deg\n",
" )\n",
" self.convs.append(conv)\n",
" self.bns.append(nn.BatchNorm1d(hidden_dim))\n",
"\n",
" self.dropout = nn.Dropout(dropout)\n",
"\n",
" # Final readout\n",
" self.readout = nn.Sequential(\n",
" nn.Linear(hidden_dim * 3, hidden_dim),\n",
" nn.ReLU(),\n",
" nn.Dropout(dropout),\n",
" nn.Linear(hidden_dim, hidden_dim // 2),\n",
" nn.ReLU(),\n",
" nn.Linear(hidden_dim // 2, 1),\n",
" )\n",
"\n",
" def forward(self, data):\n",
" x, edge_index, edge_attr, batch = (\n",
" data.x,\n",
" data.edge_index,\n",
" data.edge_attr,\n",
" data.batch,\n",
" )\n",
"\n",
" deg = degree(edge_index[0], x.size(0), dtype=torch.long).clamp(max=19)\n",
" h = self.node_emb(x) + self.deg_emb(deg)\n",
"\n",
" vn = self.virtualnode_emb(\n",
" torch.zeros(batch.max().item() + 1, dtype=torch.long, device=x.device)\n",
" )\n",
"\n",
" for conv, bn in zip(self.convs, self.bns):\n",
" h = h + vn[batch]\n",
" h = conv(h, edge_index, edge_attr)\n",
" h = bn(h)\n",
" h = F.relu(h)\n",
" h = self.dropout(h)\n",
" vn = vn + self.vn_mlp(global_mean_pool(h, batch))\n",
"\n",
" mean_pool = global_mean_pool(h, batch)\n",
" max_pool = global_max_pool(h, batch)\n",
" # add_pool = global_add_pool(h, batch)\n",
"\n",
" # graph_feats shape: [batch_size, graph_feat_dim]\n",
" if hasattr(data, 'graph_feats') and isinstance(data, torch_geometric.data.Batch):\n",
" g_feats = torch.stack([g.graph_feats for g in data.to_data_list()], dim=0).to(x.device)\n",
" else:\n",
" g_feats = data.graph_feats.unsqueeze(0).to(x.device)\n",
" g_proj = self.graph_proj(g_feats)\n",
"\n",
" final_input = torch.cat([mean_pool, max_pool, g_proj], dim=1)\n",
" return self.readout(final_input).view(-1)"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {
"id": "daTRiHSFCM9m"
},
"outputs": [],
"source": [
"from torch.amp import autocast, GradScaler\n",
"from sklearn.metrics import mean_absolute_error, mean_squared_error\n",
"\n",
"scaler_grad = GradScaler()\n",
"\n",
"def train(model, loader, optimizer, loss_fn, scaler=scaler_grad, accum_steps=8):\n",
" model.train()\n",
" total_loss = 0.0\n",
" optimizer.zero_grad()\n",
"\n",
" for i, batch in enumerate(loader):\n",
" batch = batch.to(device)\n",
" with autocast(device_type='cuda', dtype=torch.float16):\n",
" output = model(batch)\n",
" loss = loss_fn(output, batch.y.view(-1)) / accum_steps\n",
"\n",
" scaler_grad.scale(loss).backward()\n",
"\n",
" if (i + 1) % accum_steps == 0 or (i + 1 == len(loader)):\n",
" scaler.step(optimizer)\n",
" scaler.update()\n",
" optimizer.zero_grad()\n",
"\n",
" total_loss += loss.item() * batch.num_graphs * accum_steps\n",
"\n",
" return total_loss / len(loader.dataset)\n",
"\n",
"def compute_metrics_with_ci(trues, preds, n_boot=2000, alpha=0.05, seed=42):\n",
" trues = np.array(trues)\n",
" preds = np.array(preds)\n",
" mae = mean_absolute_error(trues, preds)\n",
" rmse = np.sqrt(mean_squared_error(trues, preds))\n",
"\n",
" rng = np.random.RandomState(seed)\n",
" mae_samples, rmse_samples = [], []\n",
" n = len(trues)\n",
" for _ in range(n_boot):\n",
" idx = rng.randint(0, n, n)\n",
" t, p = trues[idx], preds[idx]\n",
" mae_samples.append(mean_absolute_error(t, p))\n",
" rmse_samples.append(np.sqrt(mean_squared_error(t, p)))\n",
"\n",
" lower, upper = 100 * alpha / 2, 100 * (1 - alpha / 2)\n",
" mae_ci = (np.percentile(mae_samples, lower), np.percentile(mae_samples, upper))\n",
" rmse_ci = (np.percentile(rmse_samples, lower), np.percentile(rmse_samples, upper))\n",
" return {'mae': mae, 'mae_ci': mae_ci, 'rmse': rmse, 'rmse_ci': rmse_ci}\n",
"\n",
"def evaluate(model, loader):\n",
" model.eval()\n",
" preds, trues = [], []\n",
" with torch.no_grad(), autocast(device_type='cuda', dtype=torch.float16):\n",
" for batch in loader:\n",
" batch = batch.to(device)\n",
" out = model(batch)\n",
" preds.append(out.view(-1))\n",
" trues.append(batch.y.view(-1))\n",
" preds = torch.cat(preds)\n",
" trues = torch.cat(trues)\n",
" return torch.sqrt(torch.mean((preds - trues) ** 2)).item()\n",
"\n",
"def evaluate_with_ci(model, loader):\n",
" model.eval()\n",
" preds, trues = [], []\n",
" with torch.no_grad(), autocast(device_type='cuda', dtype=torch.float16):\n",
" for batch in loader:\n",
" batch = batch.to(device)\n",
" out = model(batch)\n",
" y_true = batch.y.view(-1).cpu().tolist()\n",
" y_pred = out.view(-1).cpu().tolist()\n",
" trues.extend(y_true)\n",
" preds.extend(y_pred)\n",
" return compute_metrics_with_ci(trues, preds)\n",
"\n"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {
"id": "b7jK5FEjCM9m"
},
"outputs": [],
"source": [
"loading = True\n",
"if loading:\n",
" train_set = torch.load('/content/drive/MyDrive/lipophil_polyatomic_data/polyatomic_data_lipophil.pt', weights_only=False)\n",
" test_set = torch.load('/content/drive/MyDrive/lipophil_polyatomic_data/polyatomic_test_data_lipophil.pt', weights_only=False)"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {
"id": "aT9Ljr-SgTsE"
},
"outputs": [],
"source": [
"import warnings\n",
"warnings.filterwarnings(\"ignore\", message=\".*An output with one or more elements was resized.*\")\n",
"warnings.filterwarnings(\"ignore\", message=\".*FutureWarning:.*\")\n",
"warnings.filterwarnings(\"ignore\", category=UserWarning, module=\"torch.optim.lr_scheduler\")"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {
"id": "_ZBbQn-98XEu"
},
"outputs": [],
"source": [
"from torch_geometric.loader import DataLoader\n",
"\n",
"train_loader = DataLoader(train_set, batch_size=128, shuffle=True, num_workers=8, pin_memory=True)\n",
"test_loader = DataLoader(test_set, batch_size=128)"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {
"id": "XchdS-ZoCM9m"
},
"outputs": [],
"source": [
"node_feat_dim = train_set[0].x.shape[1]\n",
"edge_feat_dim = train_set[0].edge_attr.shape[1]\n",
"graph_feat_dim = train_set[0].graph_feats.shape[0]\n",
"\n",
"from torch_geometric.utils import degree\n",
"\n",
"deg = torch.zeros(128, dtype=torch.long)\n",
"\n",
"for data in train_loader:\n",
" d = degree(data.edge_index[1], num_nodes=data.num_nodes, dtype=torch.long)\n",
" bc = torch.bincount(d, minlength=deg.size(0))\n",
" if bc.size(0) > deg.size(0):\n",
" new_deg = torch.zeros(bc.size(0), dtype=torch.long)\n",
" new_deg[:deg.size(0)] = deg\n",
" deg = new_deg\n",
" deg += bc\n",
"\n",
"model = PolyatomicNet(\n",
" node_feat_dim=node_feat_dim,\n",
" edge_feat_dim=edge_feat_dim,\n",
" graph_feat_dim=graph_feat_dim,\n",
" deg=deg,\n",
")\n",
"\n",
"input_dim = train_set[0].x.size(1)\n",
"optimizer = torch.optim.Adam(model.parameters(), lr=1e-3)\n",
"scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(\n",
" optimizer, mode='min', patience=10, factor=0.5, verbose=True\n",
")\n",
"loss_fn = nn.SmoothL1Loss(beta=0.5)\n",
"\n",
"device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
"model = model.to(device)"
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "Sp8U1CA2JmoL",
"outputId": "2239fd38-c080-43d6-ac3c-8a1a1a8e1421"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Epoch 01 | Train Loss: 0.5943 | Test MAE: 0.8046 (95% CI [0.7590, 0.8513]) | Test RMSE: 1.0218 (95% CI [0.9617, 1.0806])\n",
"Epoch 02 | Train Loss: 0.5771 | Test MAE: 0.8002 (95% CI [0.7539, 0.8467]) | Test RMSE: 1.0185 (95% CI [0.9580, 1.0792])\n",
"Epoch 03 | Train Loss: 0.5605 | Test MAE: 0.7535 (95% CI [0.7103, 0.7986]) | Test RMSE: 0.9599 (95% CI [0.9031, 1.0171])\n",
"Epoch 04 | Train Loss: 0.5315 | Test MAE: 0.7404 (95% CI [0.7016, 0.7805]) | Test RMSE: 0.9239 (95% CI [0.8766, 0.9719])\n",
"Epoch 05 | Train Loss: 0.5032 | Test MAE: 0.8859 (95% CI [0.8412, 0.9308]) | Test RMSE: 1.0662 (95% CI [1.0193, 1.1121])\n",
"Epoch 06 | Train Loss: 0.5014 | Test MAE: 0.7826 (95% CI [0.7362, 0.8297]) | Test RMSE: 1.0081 (95% CI [0.9490, 1.0657])\n",
"Epoch 07 | Train Loss: 0.4802 | Test MAE: 0.7119 (95% CI [0.6728, 0.7523]) | Test RMSE: 0.8873 (95% CI [0.8403, 0.9350])\n",
"Epoch 08 | Train Loss: 0.4727 | Test MAE: 0.7104 (95% CI [0.6703, 0.7509]) | Test RMSE: 0.8859 (95% CI [0.8378, 0.9337])\n",
"Epoch 09 | Train Loss: 0.4673 | Test MAE: 0.7104 (95% CI [0.6729, 0.7490]) | Test RMSE: 0.8803 (95% CI [0.8345, 0.9273])\n",
"Epoch 10 | Train Loss: 0.4557 | Test MAE: 0.7119 (95% CI [0.6721, 0.7506]) | Test RMSE: 0.8815 (95% CI [0.8354, 0.9291])\n",
"Epoch 11 | Train Loss: 0.4598 | Test MAE: 0.7039 (95% CI [0.6664, 0.7417]) | Test RMSE: 0.8657 (95% CI [0.8221, 0.9104])\n",
"Epoch 12 | Train Loss: 0.4408 | Test MAE: 0.7127 (95% CI [0.6723, 0.7565]) | Test RMSE: 0.9122 (95% CI [0.8589, 0.9676])\n",
"Epoch 13 | Train Loss: 0.4392 | Test MAE: 0.6718 (95% CI [0.6346, 0.7091]) | Test RMSE: 0.8371 (95% CI [0.7915, 0.8842])\n",
"Epoch 14 | Train Loss: 0.4239 | Test MAE: 0.9362 (95% CI [0.8893, 0.9794]) | Test RMSE: 1.1119 (95% CI [1.0635, 1.1551])\n",
"Epoch 15 | Train Loss: 0.4135 | Test MAE: 0.8397 (95% CI [0.7930, 0.8918]) | Test RMSE: 1.0707 (95% CI [1.0128, 1.1294])\n",
"Epoch 16 | Train Loss: 0.3985 | Test MAE: 0.7541 (95% CI [0.7098, 0.7995]) | Test RMSE: 0.9647 (95% CI [0.9087, 1.0179])\n",
"Epoch 17 | Train Loss: 0.4162 | Test MAE: 0.7120 (95% CI [0.6718, 0.7565]) | Test RMSE: 0.9148 (95% CI [0.8641, 0.9684])\n",
"Epoch 18 | Train Loss: 0.3940 | Test MAE: 0.6237 (95% CI [0.5877, 0.6614]) | Test RMSE: 0.7984 (95% CI [0.7517, 0.8465])\n",
"Epoch 19 | Train Loss: 0.3880 | Test MAE: 0.6442 (95% CI [0.6067, 0.6835]) | Test RMSE: 0.8313 (95% CI [0.7835, 0.8805])\n",
"Epoch 20 | Train Loss: 0.3973 | Test MAE: 0.6318 (95% CI [0.5954, 0.6711]) | Test RMSE: 0.8211 (95% CI [0.7725, 0.8727])\n",
"Epoch 21 | Train Loss: 0.3763 | Test MAE: 0.5911 (95% CI [0.5569, 0.6279]) | Test RMSE: 0.7676 (95% CI [0.7202, 0.8155])\n",
"Epoch 22 | Train Loss: 0.3671 | Test MAE: 0.6434 (95% CI [0.6052, 0.6847]) | Test RMSE: 0.8352 (95% CI [0.7863, 0.8878])\n",
"Epoch 23 | Train Loss: 0.3604 | Test MAE: 0.6489 (95% CI [0.6116, 0.6905]) | Test RMSE: 0.8388 (95% CI [0.7894, 0.8894])\n",
"Epoch 24 | Train Loss: 0.3413 | Test MAE: 0.6602 (95% CI [0.6228, 0.6978]) | Test RMSE: 0.8235 (95% CI [0.7787, 0.8676])\n",
"Epoch 25 | Train Loss: 0.3413 | Test MAE: 0.6013 (95% CI [0.5652, 0.6375]) | Test RMSE: 0.7824 (95% CI [0.7354, 0.8300])\n",
"Epoch 26 | Train Loss: 0.3379 | Test MAE: 0.6744 (95% CI [0.6387, 0.7120]) | Test RMSE: 0.8354 (95% CI [0.7897, 0.8826])\n",
"Epoch 27 | Train Loss: 0.3287 | Test MAE: 0.6842 (95% CI [0.6454, 0.7249]) | Test RMSE: 0.8793 (95% CI [0.8273, 0.9300])\n",
"Epoch 28 | Train Loss: 0.3380 | Test MAE: 0.5912 (95% CI [0.5562, 0.6268]) | Test RMSE: 0.7654 (95% CI [0.7200, 0.8120])\n",
"Epoch 29 | Train Loss: 0.3172 | Test MAE: 0.6528 (95% CI [0.6163, 0.6938]) | Test RMSE: 0.8352 (95% CI [0.7907, 0.8830])\n",
"Epoch 30 | Train Loss: 0.3174 | Test MAE: 0.9994 (95% CI [0.9520, 1.0473]) | Test RMSE: 1.1705 (95% CI [1.1227, 1.2189])\n",
"Epoch 31 | Train Loss: 0.3206 | Test MAE: 0.7928 (95% CI [0.7496, 0.8394]) | Test RMSE: 0.9942 (95% CI [0.9428, 1.0472])\n",
"Epoch 32 | Train Loss: 0.3183 | Test MAE: 0.5625 (95% CI [0.5288, 0.5980]) | Test RMSE: 0.7300 (95% CI [0.6829, 0.7789])\n",
"Epoch 33 | Train Loss: 0.3149 | Test MAE: 0.6162 (95% CI [0.5791, 0.6523]) | Test RMSE: 0.8013 (95% CI [0.7536, 0.8487])\n",
"Epoch 34 | Train Loss: 0.3104 | Test MAE: 0.6204 (95% CI [0.5827, 0.6592]) | Test RMSE: 0.7950 (95% CI [0.7504, 0.8436])\n",
"Epoch 35 | Train Loss: 0.3050 | Test MAE: 0.5997 (95% CI [0.5628, 0.6359]) | Test RMSE: 0.7868 (95% CI [0.7366, 0.8359])\n",
"Epoch 36 | Train Loss: 0.3138 | Test MAE: 0.5355 (95% CI [0.5031, 0.5705]) | Test RMSE: 0.7058 (95% CI [0.6588, 0.7543])\n",
"Epoch 37 | Train Loss: 0.2936 | Test MAE: 0.7361 (95% CI [0.6955, 0.7777]) | Test RMSE: 0.9210 (95% CI [0.8731, 0.9689])\n",
"Epoch 38 | Train Loss: 0.2766 | Test MAE: 0.5961 (95% CI [0.5623, 0.6318]) | Test RMSE: 0.7625 (95% CI [0.7154, 0.8112])\n",
"Epoch 39 | Train Loss: 0.2824 | Test MAE: 0.5254 (95% CI [0.4937, 0.5598]) | Test RMSE: 0.6938 (95% CI [0.6469, 0.7405])\n",
"Epoch 40 | Train Loss: 0.2883 | Test MAE: 0.6495 (95% CI [0.6102, 0.6912]) | Test RMSE: 0.8463 (95% CI [0.7953, 0.8971])\n",
"Epoch 41 | Train Loss: 0.2879 | Test MAE: 0.7258 (95% CI [0.6879, 0.7659]) | Test RMSE: 0.8936 (95% CI [0.8490, 0.9398])\n",
"Epoch 42 | Train Loss: 0.2796 | Test MAE: 0.6922 (95% CI [0.6523, 0.7335]) | Test RMSE: 0.8885 (95% CI [0.8383, 0.9384])\n",
"Epoch 43 | Train Loss: 0.2775 | Test MAE: 0.6467 (95% CI [0.6087, 0.6852]) | Test RMSE: 0.8160 (95% CI [0.7681, 0.8654])\n",
"Epoch 44 | Train Loss: 0.2754 | Test MAE: 0.6010 (95% CI [0.5643, 0.6381]) | Test RMSE: 0.7886 (95% CI [0.7390, 0.8376])\n",
"Epoch 45 | Train Loss: 0.2635 | Test MAE: 0.5278 (95% CI [0.4958, 0.5620]) | Test RMSE: 0.6978 (95% CI [0.6502, 0.7458])\n",
"Epoch 46 | Train Loss: 0.2594 | Test MAE: 0.5359 (95% CI [0.5006, 0.5720]) | Test RMSE: 0.7229 (95% CI [0.6721, 0.7738])\n",
"Epoch 47 | Train Loss: 0.2599 | Test MAE: 0.5426 (95% CI [0.5093, 0.5773]) | Test RMSE: 0.7057 (95% CI [0.6610, 0.7515])\n",
"Epoch 48 | Train Loss: 0.2649 | Test MAE: 0.5860 (95% CI [0.5506, 0.6228]) | Test RMSE: 0.7665 (95% CI [0.7181, 0.8148])\n",
"Epoch 49 | Train Loss: 0.2507 | Test MAE: 0.5366 (95% CI [0.5050, 0.5700]) | Test RMSE: 0.7033 (95% CI [0.6608, 0.7471])\n"
]
}
],
"source": [
"history = {}\n",
"N = 50\n",
"for epoch in range(1, N):\n",
" tr_loss = train(model, train_loader, optimizer, loss_fn)\n",
" metrics = evaluate_with_ci(model, test_loader)\n",
" print(f\"Epoch {epoch:02d} | Train Loss: {tr_loss:.4f} | \"\n",
" f\"Test MAE: {metrics['mae']:.4f} (95% CI [{metrics['mae_ci'][0]:.4f}, {metrics['mae_ci'][1]:.4f}]) | \"\n",
" f\"Test RMSE: {metrics['rmse']:.4f} (95% CI [{metrics['rmse_ci'][0]:.4f}, {metrics['rmse_ci'][1]:.4f}])\")\n",
" history[epoch] = metrics\n",
" scheduler.step(metrics['mae'])"
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "bCUfyuUlKjJb",
"outputId": "58469f70-ea2a-4a04-951f-7ee9f902dba7"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"********************\n",
"Test MAE: 0.5366 (95% CI [0.5050, 0.5700])\n",
"Test RMSE: 0.7033 (95% CI [0.6608, 0.7471])\n",
"********************\n"
]
}
],
"source": [
"final = history[N-1]\n",
"print(\"*\"*20)\n",
"print(f\"Test MAE: {final['mae']:.4f} (95% CI [{final['mae_ci'][0]:.4f}, {final['mae_ci'][1]:.4f}])\")\n",
"print(f\"Test RMSE: {final['rmse']:.4f} (95% CI [{final['rmse_ci'][0]:.4f}, {final['rmse_ci'][1]:.4f}])\")\n",
"print(\"*\"*20)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"id": "p12w2xn4Lb-U"
},
"outputs": [],
"source": []
}
],
"metadata": {
"accelerator": "GPU",
"colab": {
"gpuType": "T4",
"machine_shape": "hm",
"provenance": []
},
"kernelspec": {
"display_name": "Python 3",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.12.11"
}
},
"nbformat": 4,
"nbformat_minor": 0
}
|