init commit: загрузка и предобработка данных

core/nn/data.py

@@ -0,0 +1,23 @@
+import numpy as np
+import torch
+
+
+class TableDataset(torch.utils.data.Dataset):
+
+    def __init__(self, x: np.ndarray, y: np.ndarray) -> None:
+        """Простой датасет из объектов numpy
+
+        Args:
+            x (np.ndarray): обучающие объекты
+            y (np.ndarray): метки
+        """
+        super().__init__()
+        self.map_labels = None
+        self.x = x
+        self.y = y
+
+    def __len__(self):
+        return len(self.x)
+
+    def __getitem__(self, idx) -> tuple[torch.Tensor, torch.Tensor]:
+        return torch.FloatTensor(self.x[idx]), torch.FloatTensor(self.y[idx])

core/nn/provider.py

@@ -0,0 +1,53 @@
+import pytorch_lightning as pl
+import torch
+
+
+class NetProvider(pl.LightningModule):
+    def __init__(self, net: torch.nn.Module, lr: float, criteria: torch.nn.Module, metrics: list[callable]) -> None:
+        """Базовая обертка для сетей
+
+        Args:
+            net (torch.nn.Module): сеть - модуль pyTorch
+            lr (float): темп обучения
+            criteria (torch.nn.Module): функция ошибки
+            metrics (list[callable]): список метрик - функций, принимающих (preds: np.ndarray, true_labels: np.ndarray)
+        """
+        super().__init__()
+        self.lr = lr
+        self.criteria = criteria
+        self.net = net
+        self.metrics = metrics
+
+    def forward(self, x):
+        return self.net(x)
+
+    def configure_optimizers(self):
+        return torch.optim.AdamW(self.parameters(), lr=self.lr)
+
+    def training_step(self, batch, batch_idx):
+        data, labels = batch
+        preds = self.forward(data)
+        loss = self.criteria(preds, labels)
+        self.log("train/loss", loss, on_step=True, on_epoch=True)
+        for metric in self.metrics:
+            self.log(
+                f"train/{metric.__name__}",
+                metric(preds.cpu().detach().numpy(), labels.cpu().detach().numpy()),
+                on_step=True,
+                on_epoch=True,
+            )
+        return loss
+
+    def validation_step(self, batch, batch_idx):
+        data, labels = batch
+        preds = self.forward(data)
+        loss = self.criteria(preds, labels)
+        self.log("val/loss", loss, on_step=True, on_epoch=True)
+        for metric in self.metrics:
+            self.log(
+                f"val/{metric.__name__}",
+                metric(preds.cpu().detach().numpy(), labels.cpu().detach().numpy()),
+                on_step=True,
+                on_epoch=True,
+            )
+        return loss