first commit

.gitignore

@@ -0,0 +1,4 @@
+models/
+Datasets/
+test_images/
+train_logs/

MyDataExamplePL.py

@@ -0,0 +1,91 @@
+# %% Импорт библиотек
+
+import kagglehub
+import splitfolders
+import pytorch_lightning as pl
+
+from torch.utils.data import DataLoader
+from torchvision.datasets import ImageFolder
+from torchvision.transforms import v2 as transforms
+
+from pathlib import Path
+from typing import Optional, Callable, Any
+
+# %% Скачивание набора данных
+
+path = kagglehub.dataset_download("jehanbhathena/weather-dataset",
+                                  output_dir="./Datasets",
+                                  force_download=False)
+
+# %% Разделение набора данных
+
+splitfolders.ratio("./Datasets/weather-dataset",
+                   output="./Datasets/weather-dataset-split",
+                   ratio=(.8, .1, .1),
+                   group_prefix=None,
+                   move=False)
+
+# %% Пример модуля для загрузки данных
+
+TransformType = Optional[Callable[..., Any]]
+
+class MyImageDataset(pl.LightningDataModule):
+    def __init__(self,
+                 batch_size: int = 1,
+                 train_transform: TransformType = None,
+                 val_transform: TransformType = None,
+                 img_root: str = ""
+                 ):
+        
+        super().__init__()
+        
+        self.batch_size = batch_size
+        self.train_transform = train_transform
+        self.val_transform = val_transform
+        self.img_root = img_root
+        
+    def setup(self, stage=None):
+        self.train_set = ImageFolder(self.img_root / Path("train"), 
+                                     transform=self.train_transform)
+        self.val_set = ImageFolder(self.img_root / Path("val"), 
+                                     transform=self.val_transform)
+        self.test_set = ImageFolder(self.img_root / Path("test"), 
+                                     transform=self.val_transform)
+        
+    def train_dataloader(self):
+        return DataLoader(self.train_set, batch_size=self.batch_size,
+                          shuffle=True, num_workers=2)
+    
+    def val_dataloader(self):
+        return DataLoader(self.val_set, batch_size=self.batch_size,
+                          shuffle=False, num_workers=2)
+    
+    def test_dataloader(self):
+        return DataLoader(self.test_set, batch_size=self.batch_size,
+                          shuffle=False, num_workers=2)
+
+# %% Создание трансформаций
+
+transform_train = transforms.Compose([
+    transforms.Resize((224, 224)),
+    transforms.RandomHorizontalFlip(p=.5),
+    transforms.ToTensor(),
+    transforms.Normalize(mean=[0.485, 0.456, 0.406],
+                         std=[0.229, 0.224, 0.225])     
+])
+
+transform_test = transforms.Compose([
+    transforms.Resize((224, 224)),
+    transforms.ToTensor(),
+    transforms.Normalize(mean=[0.485, 0.456, 0.406],
+                         std=[0.229, 0.224, 0.225])     
+])
+
+# %% Создание экземляра класса и вывод объектов класса
+
+dm = MyImageDataset(train_transform=transform_train,
+                    val_transform=transform_test,
+                    img_root="./Datasets/weather-dataset-split")
+dm.setup()
+
+print({k:v for k, v in enumerate(dm.test_set.classes)})

MyNNExamplePL.py

@@ -0,0 +1,229 @@
+# %% Импорт библиотек
+
+import torch
+import torch.nn as nn
+import pytorch_lightning as pl
+
+from torch.utils.data import DataLoader
+from torchmetrics import MetricCollection
+from torchvision import datasets, transforms
+from torchmetrics.classification import Accuracy, Precision, Recall, F1Score
+
+# Автоматическое логирование результатов
+from pytorch_lightning.loggers import WandbLogger, TensorBoardLogger, CSVLogger
+# сохранение весов
+from pytorch_lightning.callbacks.model_checkpoint import ModelCheckpoint
+# автоматическое отслеживание lr
+from pytorch_lightning.callbacks import LearningRateMonitor
+# ранние остановки
+from pytorch_lightning.callbacks.early_stopping import EarlyStopping
+
+# %% Гиперпараметры
+
+num_epoch = 8
+batch_size = 64
+
+# %% Создание модели PyTorch
+
+class MyTorchCNNModel(nn.Module):
+    def __init__(self, 
+                 output_size: int = 10):
+        super().__init__()
+        
+        self.output_size = output_size
+        
+        self.feature = nn.Sequential(
+            nn.LazyConv2d(64, kernel_size=3, stride=1, padding=1),
+            nn.BatchNorm2d(64),
+            nn.ReLU(),
+            nn.Conv2d(64, 128, kernel_size=3, stride=1, padding=1),
+            nn.BatchNorm2d(128),
+            nn.ReLU(),
+            nn.Conv2d(128, 256, kernel_size=3, stride=1, padding=1),
+            nn.BatchNorm2d(256),
+            nn.ReLU(),
+            nn.MaxPool2d(kernel_size=2)
+        )
+        
+        self.head = nn.Conv2d(256, self.output_size, kernel_size=1)
+        self.pool = nn.AdaptiveAvgPool2d((1, 1))
+        
+        
+    def forward(self, image):
+        x = self.feature(image)
+        x = self.head(x)
+        x = self.pool(x)
+        
+        return x.flatten(1)
+
+# %% Создание модели PyTorch Lightning
+
+class MyNNExample(pl.LightningModule):
+    def __init__(self,
+                 model: nn.Module,
+                 eta=3e-4):
+        super().__init__()
+        
+        self.save_hyperparameters(ignore=["model"])
+        # device = "cuda" if torch.cuda.is_available() else "cpu"
+        self.model = model
+        
+        self.criterion = nn.CrossEntropyLoss()
+        self.train_metrics = MetricCollection({
+            "accuracy": Accuracy(task="multiclass", 
+                                 num_classes=model.output_size),
+            "precision": Precision(task="multiclass", 
+                                   num_classes=model.output_size, 
+                                   average="macro"),    
+            "recall": Recall(task="multiclass", 
+                             num_classes=model.output_size, 
+                             average="macro"),
+            "f1": F1Score(task="multiclass", 
+                          num_classes=model.output_size, 
+                          average="macro")    
+        })
+        
+        self.val_metrics = self.train_metrics.clone()
+        self.test_metrics = self.train_metrics.clone()
+        
+        
+    def forward(self, image):
+        return self.model(image)
+        
+
+    def _shared_step(self,
+                    sample,
+                    stage,
+                    metrics):
+        x, y = sample
+        logits = self.forward(x.to(torch.float32))
+        preds = torch.argmax(logits, 1)
+        loss = self.criterion(logits, y.to(torch.int64))
+        
+        metrics.update(preds, y)
+        self.log(f"{stage}_loss", loss, 
+                 prog_bar=True, on_epoch=True, on_step=False)
+        
+        return loss
+    
+    def _log_metrics(self, stage, metrics):
+        computed = metrics.compute()
+        
+        for name, value in computed.items():
+            self.log(f"{stage}_{name}", value, prog_bar=True)
+        
+        metrics.reset()
+    
+    def training_step(self, batch, batch_idx):
+        return self._shared_step(batch, "train", self.train_metrics)
+    
+    def on_train_epoch_end(self):
+        self._log_metrics("train", self.train_metrics)
+    
+    def validation_step(self, batch, batch_idx):
+        return self._shared_step(batch, "val", self.val_metrics)
+    
+    def on_validation_epoch_end(self):
+        self._log_metrics("val", self.val_metrics)
+    
+    def test_step(self, batch, batch_idx):
+        return self._shared_step(batch, "test", self.test_metrics)
+    
+    def on_test_epoch_end(self):
+       self._log_metrics("test", self.test_metrics)
+    
+    def configure_optimizers(self):
+        optimizer = torch.optim.Adam(self.parameters(), lr=self.hparams.eta)
+        scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(
+            optimizer, patience=5, mode="min"
+        )
+        
+        return {
+            "optimizer": optimizer,
+            "lr_scheduler": {
+                "scheduler": scheduler,
+                "monitor": "val_loss",
+                "interval": "epoch",
+            }
+        }
+
+# %% Загрузка набора данных
+
+image_transforms = transforms.Compose([
+    transforms.ToTensor()
+])
+
+train_dataset = datasets.MNIST(
+    root="./Datasets",
+    train=True,
+    download=False, # Если скачен, то убрать
+    transform=image_transforms
+)
+
+test_dataset = datasets.MNIST(
+    root="./Datasets",
+    train=False,
+    download=False, # Если скачен, то убрать
+    transform=image_transforms    
+)
+
+train_loader = DataLoader(dataset=train_dataset,
+                          batch_size=batch_size,
+                          shuffle=True)
+
+test_loader = DataLoader(dataset=test_dataset,
+                         batch_size=batch_size,
+                         shuffle=False)
+
+# %% Запуск обучения
+
+model = MyNNExample(model=MyTorchCNNModel())
+callbacks = [
+    ModelCheckpoint(dirpath="models",
+                    filename="{epoch}_{val_accuracy:.4f}_{val_loss:.4f}",
+                    auto_insert_metric_name=False,
+                    save_last=True,
+                    save_top_k=1,
+                    monitor="val_loss",
+                    mode="min"),
+    LearningRateMonitor(logging_interval="step"),
+    EarlyStopping(monitor="val_loss",
+                  min_delta=2e-4,
+                  patience=10,
+                  verbose=False,
+                  mode="min")
+]
+
+trainer = pl.Trainer(
+    max_epochs=num_epoch,
+    accelerator="gpu",
+    devices=1,
+    logger=CSVLogger("./train_logs", name="mnist_csv", 
+                     version=None, flush_logs_every_n_steps=100),
+    callbacks=callbacks
+)
+
+trainer.fit(model=model, 
+            train_dataloaders=train_loader,
+            val_dataloaders=test_loader)
+        
+# %% Инференс модели
+        
+checkpoint_path = "./models/6_0.9460_0.1643.ckpt"
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+model = MyNNExample.load_from_checkpoint(checkpoint_path,
+                                         model=MyTorchCNNModel(output_size=10))
+model = model.to(device)
+model.eval()
+
+image, true_label = test_dataset[42]
+image_batch = image.unsqueeze(0).to(device)
+
+with torch.no_grad():
+    logits = model(image_batch)
+    probs = torch.softmax(logits, dim=1)
+    pred_label = logits.argmax(dim=1).item()
+    
+print(f"Предсказание: {pred_label}")
+print(f"Истинная метка: {true_label}")
+print(f"Уверенность: {probs[0, pred_label].item():.2%}")

README.md

@@ -0,0 +1,19 @@
+Установка с созданием новой среды:
+```bash
+conda env create -f environment.yml
+```
+
+Установка в уже созданную среду:
+```bash
+conda env update --file environment.yml --prune
+```
+
+Для переноса в другую ОС:
+```bash
+conda env export --from-history > environment.yml
+```
+
+Просто сохранить зависимости:
+```bash
+conda env export > environment.yml
+```

SimpleMLP.py

@@ -0,0 +1,37 @@
+# %% Импорт библиотек
+import torch
+import torch.nn as nn
+
+# %% Входные переменные
+input_ = 32*32
+hidden_1 = 5
+hidden_2 = 12
+output = 2
+learning_rate = 1e-3
+
+weights_1 = torch.rand(size=(input_, hidden_1), requires_grad=True)
+weights_2 = torch.rand(size=(hidden_1, hidden_2), requires_grad=True)
+weights_3 = torch.rand(size=(hidden_2, output), requires_grad=True)
+
+random_input = torch.rand(1, input_)
+
+true_label = torch.zeros(1, output)
+true_label[0][0] = 1
+# %% Цикл обучения
+criterion = nn.CrossEntropyLoss()
+optimizer = torch.optim.SGD([weights_1, weights_2, weights_3], lr=learning_rate)
+
+hidden_layer_1 = random_input @ weights_1
+hidden_layer_2 = hidden_layer_1 @ weights_2
+output_layer = hidden_layer_2 @ weights_3
+
+loss = criterion(output_layer, true_label)
+print(loss)
+
+optimizer.zero_grad()
+loss.backward()
+optimizer.step()
+
+print(weights_1._grad)
+print(weights_2._grad)
+print(weights_3._grad)

environment.yml

@@ -0,0 +1,427 @@
+name: CV
+channels:
+  - defaults
+  - conda-forge
+dependencies:
+  - _libgcc_mutex=0.1=main
+  - _openmp_mutex=5.1=52_gnu
+  - aiodns=3.6.1=py312h06a4308_0
+  - aiohappyeyeballs=2.6.2=py312h06a4308_0
+  - aiohttp=3.14.3=py312h47b2149_0
+  - aiosignal=1.4.0=py312h06a4308_0
+  - alabaster=1.0.0=py312h06a4308_0
+  - alsa-lib=1.2.15.3=h47b2149_0
+  - anyio=4.12.1=py312h06a4308_0
+  - aom=3.13.2=h664349e_0
+  - argon2-cffi=25.1.0=py312h06a4308_0
+  - argon2-cffi-bindings=25.1.0=py312hee96239_0
+  - arrow=1.4.0=py312h06a4308_0
+  - astroid=4.0.3=py312h06a4308_0
+  - asttokens=3.0.1=py312h06a4308_0
+  - async-lru=2.3.0=py312h06a4308_0
+  - asyncssh=2.24.0=py312h06a4308_0
+  - attrs=26.1.0=py312h0c820a0_0
+  - autopep8=2.0.4=pyhd3eb1b0_0
+  - babel=2.18.0=py312h06a4308_0
+  - bcrypt=5.0.0=py312h498d7c9_0
+  - beartype=0.22.9=py312h06a4308_0
+  - beautifulsoup4=4.15.0=py312h06a4308_0
+  - binaryornot=0.6.0=py312h06a4308_0
+  - black=26.5.1=py312h06a4308_0
+  - blas=1.0=openblas
+  - bleach=6.4.0=py312h06a4308_0
+  - brotlicffi=1.2.0.1=py312h7354ed3_0
+  - bzip2=1.0.8=h5eee18b_6
+  - c-ares=1.34.7=hd44998d_0
+  - ca-certificates=2026.8.13=h06a4308_0
+  - cairo=1.18.4=h44eff21_0
+  - cattrs=26.1.0=py312h06a4308_0
+  - certifi=2026.7.22=py312h06a4308_0
+  - cffi=2.1.1=py312h3b52fac_0
+  - chardet=5.2.0=py312h06a4308_1
+  - charset-normalizer=3.4.7=py312h06a4308_0
+  - click=8.5.0=py312h06a4308_0
+  - cloudpickle=3.1.2=py312h06a4308_0
+  - comm=0.2.3=py312h06a4308_0
+  - cookiecutter=2.7.1=py312h06a4308_0
+  - cpython=3.12.14=py312hd8ed1ab_3
+  - cryptography=50.0.0=py312hef7b95b_0
+  - cuda-crt-tools=12.9.86=ha770c72_2
+  - cuda-cudart=12.9.79=h5888daf_0
+  - cuda-cudart_linux-64=12.9.79=h3f2d84a_0
+  - cuda-cuobjdump=12.9.82=hffce074_1
+  - cuda-cupti=12.9.79=h676940d_2
+  - cuda-nvcc-tools=12.9.86=he02047a_2
+  - cuda-nvdisasm=12.9.88=hffce074_1
+  - cuda-nvrtc=12.9.86=hecca717_1
+  - cuda-nvtx=12.9.79=hecca717_1
+  - cuda-nvvm-tools=12.9.86=h4bc722e_2
+  - cuda-version=12.9=h4f385c5_3
+  - cudnn=9.25.1.1=ha9dd1bd_0
+  - cusparselt=0.7.0.0=hcd2ec93_0
+  - cyrus-sasl=2.1.28=h2f687bf_5
+  - dav1d=1.5.3=h12c9f22_1
+  - dbus=1.16.2=h5bd4931_0
+  - debugpy=1.8.21=py312h7354ed3_0
+  - decorator=5.3.1=py312h06a4308_0
+  - defusedxml=0.7.1=pyhd3eb1b0_0
+  - diff-match-patch=20200713=pyhd3eb1b0_0
+  - dill=0.4.1=py312h06a4308_0
+  - docstring-to-markdown=0.17=py312h06a4308_0
+  - docutils=0.21.2=py312h06a4308_1
+  - executing=2.2.1=py312h06a4308_1
+  - expat=2.8.4=h7354ed3_0
+  - filelock=3.32.5=pyhd8ed1ab_0
+  - flake8=7.1.1=py312h06a4308_0
+  - fmt=12.1.0=h44f220b_0
+  - fontconfig=2.15.0=h2c49b7f_0
+  - freetype=2.14.1=hf5b9546_0
+  - fribidi=1.0.16=h10f3c28_1
+  - frozenlist=1.8.0=py312hbdd6827_0
+  - fsspec=2026.7.0=pyhd8ed1ab_0
+  - gettext=0.25.1=h92eb808_0
+  - gettext-tools=0.25.1=h6a67909_0
+  - giflib=6.1.3=h621bd71_1
+  - gmp=6.3.0=hfd2156b_3
+  - gmpy2=2.3.1=py312h04c1168_0
+  - graphite2=1.3.15=h9ba177b_0
+  - h11=0.16.0=py312h06a4308_1
+  - harfbuzz=12.3.0=h79d275a_1
+  - httpcore=1.0.9=py312h06a4308_0
+  - httpx=0.28.1=py312h06a4308_1
+  - icu=73.1=h6a678d5_0
+  - idna=3.18=py312h06a4308_0
+  - imagesize=1.4.1=py312h06a4308_0
+  - importlib-metadata=9.0.0=py312h06a4308_0
+  - inflection=0.5.1=py312h06a4308_1
+  - intervaltree=3.2.1=pyhd3eb1b0_0
+  - ipykernel=6.31.0=py312h7040dfc_0
+  - ipython=9.15.0=py312h06a4308_0
+  - ipython_pygments_lexers=1.1.1=py312h06a4308_0
+  - isort=8.0.1=py312h06a4308_1
+  - jansson=2.15.0=hbcba0ee_0
+  - jaraco.classes=3.4.0=py312h06a4308_0
+  - jaraco.context=6.1.0=py312h06a4308_0
+  - jaraco.functools=4.4.0=py312h06a4308_0
+  - jedi=0.19.2=py312h06a4308_0
+  - jeepney=0.7.1=pyhd3eb1b0_0
+  - jellyfish=1.2.1=py312h498d7c9_0
+  - jinja2=3.1.6=pyhcf101f3_1
+  - jpeg=9f=he7e78df_1
+  - json5=0.15.0=py312h06a4308_0
+  - jsonschema=4.26.0=py312h06a4308_0
+  - jsonschema-specifications=2025.9.1=py312h06a4308_0
+  - jupyter-builder=1.2.0=py312h06a4308_0
+  - jupyter-lsp=2.3.0=py312h06a4308_0
+  - jupyter_client=8.10.0=py312h06a4308_0
+  - jupyter_core=5.9.1=py312h06a4308_0
+  - jupyter_events=0.12.1=py312h06a4308_0
+  - jupyter_server=2.20.0=py312h06a4308_0
+  - jupyter_server_terminals=0.5.4=py312h06a4308_0
+  - jupyterlab=4.6.2=py312h06a4308_1
+  - jupyterlab_pygments=0.3.0=py312h06a4308_0
+  - jupyterlab_server=2.28.0=py312h06a4308_1
+  - kagglehub=1.0.2=py312h06a4308_0
+  - kagglesdk=0.1.35=py312h06a4308_0
+  - keyring=25.7.0=py312h06a4308_0
+  - krb5=1.20.1=h143b758_1
+  - lcms2=2.19=he283960_0
+  - ld_impl_linux-64=2.44=h9e0c5a2_3
+  - lerc=4.1.0=h7354ed3_2
+  - libabseil=20260107.0=cxx17_h6199ee8_0
+  - libasprintf=0.25.1=hf2ab22a_0
+  - libasprintf-devel=0.25.1=hf2ab22a_0
+  - libavif=1.3.0=h2b90b00_1
+  - libbrotlicommon=1.2.0=h32cd6e7_0
+  - libbrotlidec=1.2.0=ha2c5f68_0
+  - libbrotlienc=1.2.0=h2e96acb_0
+  - libcap=2.78=h084b8d7_1
+  - libcublas=12.9.2.10=h676940d_0
+  - libcudnn=9.25.1.1=hfd19c1e_0
+  - libcudnn-dev=9.25.1.1=ha9dd1bd_0
+  - libcudss0=0.4.0.2=he55f5cd_2
+  - libcufft=11.4.1.4=hecca717_1
+  - libcufile=1.14.1.1=hbc026e6_1
+  - libcups=2.4.19=h23114ae_0
+  - libcurand=10.3.10.19=h676940d_1
+  - libcurl=8.19.0=ha05a353_0
+  - libcusolver=11.7.5.82=h676940d_2
+  - libcusparse=12.5.10.65=hecca717_2
+  - libdeflate=1.22=h5eee18b_0
+  - libdrm=2.4.134=h9c74679_0
+  - libedit=3.1.20230828=h5eee18b_0
+  - libegl=1.7.0=h5eee18b_2
+  - libev=4.33=h7f8727e_1
+  - libevent=2.1.12=hdbd6064_1
+  - libexpat=2.8.4=h7354ed3_0
+  - libffi=3.4.8=h06d3fd0_3
+  - libgcc=15.2.0=h69a1729_8
+  - libgcc-ng=15.2.0=h166f726_8
+  - libgettextpo=0.25.1=hf2ab22a_0
+  - libgettextpo-devel=0.25.1=hf2ab22a_0
+  - libgfortran=15.2.0=h69a702a_20
+  - libgfortran5=15.2.0=h68bc16d_20
+  - libgl=1.7.0=h5eee18b_2
+  - libglib=2.88.3=ha16e27a_0
+  - libglvnd=1.7.0=h5eee18b_2
+  - libglx=1.7.0=h5eee18b_2
+  - libgomp=15.2.0=h4751f2c_8
+  - libiconv=1.18=h75a1612_0
+  - libidn2=2.3.8=hf80d704_0
+  - libkrb5=1.22.1=h6d2bf13_0
+  - libllvm20=20.1.8=h465586c_0
+  - libllvm22=22.1.2=hfb1d434_0
+  - liblzma=5.8.2=hb03c661_0
+  - libnghttp2=1.67.1=h697f920_0
+  - libnl=3.11.0=hb03c661_1
+  - libnvjitlink=12.9.86=hecca717_2
+  - libnvjpeg=12.4.0.76=h7354ed3_0
+  - libopenblas=0.3.34=pthreads_h94d23a6_0
+  - libopengl=1.7.0=h5eee18b_2
+  - libopenjpeg=2.5.4=hee96239_1
+  - libopus=1.6.1=h9f10d21_1
+  - libpciaccess=0.18=h5eee18b_0
+  - libpng=1.6.56=h22898a0_0
+  - libpq=17.9=h0cb448f_1
+  - libprotobuf=6.33.5=h4435b4a_0
+  - libre2-11=2025.11.05=h3356fce_1
+  - libsodium=1.0.21=h83fc4cd_1
+  - libspatialindex=2.1.0=hf7d4471_0
+  - libssh2=1.11.1=hfbabe93_1
+  - libstdcxx=15.2.0=h39759b7_8
+  - libstdcxx-ng=15.2.0=hc03a8fd_8
+  - libsystemd0=261.2=h6f4a2f1_0
+  - libtiff=4.7.1=h029b1ac_0
+  - libtirpc=1.3.7=h47b2149_0
+  - libtorch=2.12.0=gpu_cuda129_h0198de7_200
+  - libudev1=261.2=h6f4a2f1_0
+  - libunistring=1.4.2=h34b0ebb_0
+  - libuuid=1.41.5=h5eee18b_0
+  - libuv=1.52.1=h280c20c_1
+  - libvpx=1.16.0=h4b463fa_0
+  - libwebp-base=1.6.0=hb7bb969_0
+  - libxcb=1.17.0=h9b100fa_0
+  - libxkbcommon=1.9.1=h69220b7_0
+  - libxkbfile=1.1.0=hf54ef8a_2
+  - libxml2=2.13.9=h2c43086_0
+  - libxslt=1.1.43=h28b3bda_0
+  - libzlib=1.3.2=h47b2149_0
+  - lightning-utilities=0.15.3=pyhd8ed1ab_0
+  - lmdb=1.0.0=hfe55579_0
+  - lsprotocol=2025.0.0=py312h06a4308_1
+  - lz4-c=1.9.4=h7354ed3_5
+  - magma=2.10.0=hfa452ef_1
+  - markdown-it-py=4.2.0=py312h06a4308_0
+  - markupsafe=3.0.3=py312h47b2149_0
+  - matplotlib-inline=0.2.2=py312h06a4308_0
+  - mccabe=0.7.0=pyhd3eb1b0_0
+  - mdurl=0.1.2=py312h06a4308_0
+  - mesalib=25.1.5=h3583ad3_5
+  - minizip=4.2.1=h7e49174_0
+  - mistune=3.3.3=py312h06a4308_0
+  - more-itertools=11.0.2=py312h06a4308_0
+  - mpc=1.4.0=ha2cb11d_1
+  - mpfr=4.2.2=ha2cb11d_1
+  - mpmath=1.4.1=pyhd8ed1ab_0
+  - multidict=6.7.1=py312h47b2149_0
+  - mypy_extensions=1.1.0=py312h06a4308_0
+  - mysql=9.3.0=hef88997_5
+  - mysql-client=9.3.0=hf2c64df_5
+  - mysql-common=9.3.0=h9e076cb_5
+  - mysql-devel=9.3.0=h9e076cb_5
+  - mysql-libs=9.3.0=h6ecde68_5
+  - mysql-server=9.3.0=h284711e_5
+  - nbclient=0.11.0=py312h06a4308_0
+  - nbconvert=7.17.1=py312h06a4308_0
+  - nbconvert-core=7.17.1=py312h06a4308_0
+  - nbconvert-pandoc=7.17.1=py312h06a4308_0
+  - nbformat=5.11.0=py312h06a4308_0
+  - nccl=2.30.7.1=h4d09622_0
+  - ncurses=6.6=hfaaeb4e_0
+  - nest-asyncio=1.6.0=py312h06a4308_1
+  - nest-asyncio2=1.7.2=py312h06a4308_0
+  - networkx=3.6.1=pyhcf101f3_0
+  - nomkl=1.0=h5ca1d4c_0
+  - notebook-shim=0.2.4=py312h06a4308_1
+  - nspr=4.39=h6e5897e_0
+  - nss=3.125=hcf7f90d_0
+  - numpy=2.5.2=py312h35deafb_0
+  - numpy-base=2.5.2=py312h4bc27c9_0
+  - numpydoc=1.10.0=py312h06a4308_0
+  - openjpeg=2.5.4=h4e0627c_1
+  - openldap=2.6.12=h007892f_1
+  - openssl=3.6.4=h781a0a9_0
+  - optree=0.20.0=py312ha5105d2_1
+  - packaging=26.3=py312h06a4308_0
+  - pandoc=3.8=h06a4308_0
+  - pandocfilters=1.5.1=py312h06a4308_0
+  - parso=0.8.7=py312h06a4308_0
+  - pathspec=1.1.1=py312h06a4308_0
+  - pcre2=10.46=hf426167_0
+  - pexpect=4.9.0=py312h06a4308_1
+  - pickleshare=0.7.5=pyhd3eb1b0_1003
+  - pillow=12.2.0=py312haa37d31_0
+  - pip=26.2.1=pyhc872135_0
+  - pixman=0.46.4=h86ba9f7_1
+  - platformdirs=4.11.0=py312h06a4308_0
+  - pluggy=1.6.0=py312h06a4308_0
+  - prometheus_client=0.24.1=py312h06a4308_0
+  - prompt-toolkit=3.0.53=py312h06a4308_0
+  - prompt_toolkit=3.0.53=hd3eb1b0_0
+  - propcache=0.5.2=py312h47b2149_0
+  - protobuf=6.33.5=py312h7354ed3_0
+  - psutil=7.2.2=py312h47b2149_0
+  - pthread-stubs=0.3=h47b2149_2
+  - ptyprocess=0.7.0=pyhd3eb1b0_3
+  - pure_eval=0.2.3=py312h06a4308_0
+  - pybind11=3.0.4=py312h24d9097_0
+  - pybind11-abi=11=hd3eb1b0_0
+  - pybind11-global=3.0.4=py312h24d9097_0
+  - pycares=4.10.0=py312ha137689_0
+  - pycodestyle=2.12.1=py312h06a4308_0
+  - pycparser=3.0=py312h06a4308_0
+  - pydocstyle=6.3.0=py312h06a4308_1
+  - pyflakes=3.2.0=py312h06a4308_0
+  - pygithub=2.9.1=py312h06a4308_0
+  - pygments=2.20.0=py312h06a4308_0
+  - pyjwt=2.13.0=py312h06a4308_0
+  - pylint=4.0.6=py312h06a4308_0
+  - pylint-venv=3.0.3=py312h06a4308_0
+  - pyls-spyder=0.4.0=pyhd3eb1b0_0
+  - pynacl=1.6.2=py312h83fc4cd_1
+  - pyqt=6.11.0=py312h0eb9b55_0
+  - pyqt6-sip=13.11.1=py312h47b2149_0
+  - pyqtwebengine=6.11.0=py312hdc71590_0
+  - pysocks=1.7.1=py312h06a4308_1
+  - python=3.12.14=hc5f7cf0_0
+  - python-dateutil=2.9.0post0=py312h06a4308_2
+  - python-fastjsonschema=2.22.1=py312h06a4308_0
+  - python-json-logger=4.1.0=py312h06a4308_0
+  - python-lsp-black=2.0.0=py312h06a4308_1
+  - python-lsp-jsonrpc=1.1.2=pyhd3eb1b0_0
+  - python-lsp-ruff=2.3.0=py312h06a4308_0
+  - python-lsp-server=1.14.0=py312h7040dfc_1
+  - python-slugify=8.0.4=py312h06a4308_0
+  - python-tzdata=2026.3=pyhd3eb1b0_0
+  - python_abi=3.12=4_cp312
+  - pytokens=0.4.1=py312h47b2149_0
+  - pytoolconfig=1.3.1=py312h06a4308_0
+  - pytorch=2.12.0=gpu_cuda129_py312h95261ae_200
+  - pytorch-lightning=2.6.5=pyhcf101f3_0
+  - pyuca=1.2=py312h06a4308_1
+  - pyxdg=0.28=py312h06a4308_0
+  - pyyaml=6.0.3=py312h8a5da7c_1
+  - pyzmq=27.1.0=py312hcf8288c_1
+  - qdarkstyle=3.2.3=pyhd3eb1b0_0
+  - qstylizer=0.2.4=py312h06a4308_0
+  - qtawesome=1.4.2=py312h4333747_0
+  - qtbase=6.11.0=h9201cad_0
+  - qtconsole=5.7.2=py312h06a4308_0
+  - qtdeclarative=6.11.0=h89ee561_1
+  - qtpy=2.4.3=py312h06a4308_0
+  - qtsvg=6.11.0=hbecefc2_1
+  - qttools=6.11.0=h0af0a4d_1
+  - qtwayland=6.11.0=h9a62c7d_0
+  - qtwebchannel=6.11.0=h7ebffc9_1
+  - qtwebengine=6.11.0=hb54882b_0
+  - qtwebsockets=6.11.0=h03d4248_1
+  - rdma-core=65.0=h7b4f2e6_0
+  - re2=2025.11.05=h71b5c5c_1
+  - readline=8.3=hc2a1206_0
+  - referencing=0.37.0=py312h06a4308_0
+  - requests=2.34.2=py312h06a4308_0
+  - rfc3339-validator=0.1.4=py312h06a4308_0
+  - rfc3986-validator=0.1.1=py312h06a4308_0
+  - rich=15.0.0=py312h06a4308_0
+  - roman-numerals=3.1.0=py312h06a4308_0
+  - rope=1.14.0=py312h06a4308_0
+  - rpds-py=0.28.0=py312h498d7c9_0
+  - rtree=1.4.1=py312hd966e25_1
+  - ruff=0.15.1=py312h6e1b9ff_0
+  - secretstorage=3.5.0=py312h3e8c6aa_0
+  - send2trash=1.8.3=py312h06a4308_0
+  - setuptools=80.10.2=py312h06a4308_0
+  - sip=6.15.1=py312he6c97cb_0
+  - six=1.17.0=py312h06a4308_0
+  - sleef=3.9.0=ha0421bc_0
+  - snappy=1.2.2=h4bcf44c_1
+  - snowballstemmer=3.1.1=py312h06a4308_0
+  - sortedcontainers=2.4.0=pyhd3eb1b0_0
+  - soupsieve=2.8.4=py312h06a4308_0
+  - sphinx=9.1.0=py312h893fd2c_0
+  - sphinxcontrib-applehelp=2.0.0=pyhd3eb1b0_1
+  - sphinxcontrib-devhelp=2.0.0=pyhd3eb1b0_0
+  - sphinxcontrib-htmlhelp=2.1.0=pyhd3eb1b0_0
+  - sphinxcontrib-jsmath=1.0.1=pyhd3eb1b0_0
+  - sphinxcontrib-qthelp=2.0.0=pyhd3eb1b0_1
+  - sphinxcontrib-serializinghtml=2.0.0=pyhd3eb1b0_0
+  - spirv-tools=2026.1=h24d9097_0
+  - spyder=6.1.5=py312h06a4308_0
+  - spyder-kernels=3.1.6=py312h7040dfc_0
+  - sqlite=3.53.2=h795bf6d_0
+  - stack_data=0.6.3=py312h06a4308_0
+  - superqt=0.8.1=py312ha9b9ae4_0
+  - sympy=1.14.0=pyh2585a3b_106
+  - terminado=0.18.1=py312h06a4308_1
+  - text-unidecode=1.3=pyhd3eb1b0_0
+  - textdistance=4.6.3=py312he3bba80_1
+  - three-merge=0.1.1=pyhd3eb1b0_0
+  - tinycss2=1.5.1=py312h06a4308_0
+  - tk=8.6.15=h54e0aa7_0
+  - tomlkit=0.15.1=py312h06a4308_0
+  - torchao=0.17.0=cuda129_py312hd0f1f67_2
+  - torchmetrics=1.9.0=pyhd8ed1ab_0
+  - torchvision=0.27.0=cuda129py312h054a6d6_100
+  - tornado=6.5.8=py312h47b2149_0
+  - tqdm=4.70.0=pyh8f84b5b_0
+  - traitlets=5.15.0=py312h06a4308_0
+  - triton=3.7.0=cuda129py312h338dc61_1
+  - typing-extensions=4.16.0=py312h06a4308_0
+  - typing_extensions=4.16.0=py312h06a4308_0
+  - tzdata=2026c=he532380_0
+  - ujson=5.13.0=py312h7354ed3_0
+  - urllib3=2.7.0=py312h06a4308_0
+  - uvloop=0.22.1=py312h939ff40_1
+  - watchdog=6.0.0=py312h06a4308_0
+  - wayland=1.24.0=hdac8c69_0
+  - wcwidth=0.8.2=py312h06a4308_0
+  - webencodings=0.6.1=py312h06a4308_0
+  - websocket-client=1.8.0=py312h06a4308_0
+  - whatthepatch=1.0.7=py312h06a4308_0
+  - wheel=0.47.0=py312h06a4308_0
+  - wurlitzer=3.1.1=py312h06a4308_0
+  - xcb-util=0.4.1=h5eee18b_2
+  - xcb-util-cursor=0.1.5=h5eee18b_0
+  - xcb-util-image=0.4.0=h5eee18b_2
+  - xcb-util-keysyms=0.4.1=h5eee18b_0
+  - xcb-util-renderutil=0.3.10=h5eee18b_0
+  - xcb-util-wm=0.4.2=h5eee18b_0
+  - xkeyboard-config=2.48=h382ed1a_0
+  - xorg-libice=1.1.2=h9b100fa_0
+  - xorg-libsm=1.2.6=h9b100fa_0
+  - xorg-libx11=1.8.12=h9b100fa_1
+  - xorg-libxau=1.0.12=h9b100fa_0
+  - xorg-libxcomposite=0.4.6=h9b100fa_2
+  - xorg-libxdamage=1.1.6=h9b100fa_0
+  - xorg-libxdmcp=1.1.5=h9b100fa_0
+  - xorg-libxext=1.3.7=h1ce37a7_0
+  - xorg-libxfixes=6.0.1=h9b100fa_0
+  - xorg-libxi=1.8.2=h9b100fa_0
+  - xorg-libxrandr=1.5.4=h9b100fa_0
+  - xorg-libxrender=0.9.12=h9b100fa_0
+  - xorg-libxshmfence=1.3.3=h9b100fa_0
+  - xorg-libxtst=1.2.5=h9b100fa_3
+  - xorg-libxxf86vm=1.1.6=h9b100fa_0
+  - xorg-xorgproto=2024.1=h47b2149_2
+  - xz=5.8.2=h448239c_0
+  - yaml=0.2.5=h591646f_1
+  - yapf=0.43.0=py312h06a4308_0
+  - yarl=1.24.2=py312h47b2149_0
+  - zeromq=4.3.5=hf801bfb_2
+  - zipp=4.1.0=py312h06a4308_0
+  - zlib=1.3.2=h47b2149_0
+  - zstd=1.5.7=h11fc155_0
+  - pip:
+      - pyqt6-qt6==6.11.2
+      - split-folders==0.6.1
+prefix: /home/daniel/anaconda3/envs/CV

environment_win.yml

@@ -0,0 +1,13 @@
+name: CV
+channels:
+  - defaults
+dependencies:
+  - python=3.12
+  - jupyterlab
+  - spyder
+  - pytorch-lightning
+  - torchao
+  - beartype
+  - torchvision
+  - kagglehub
+prefix: /home/daniel/anaconda3/envs/CV

ffn.py

@@ -0,0 +1,189 @@
+# %% Импорт библиотек
+
+import tqdm
+import torch
+import torch.nn as nn
+
+from torch.utils.data import DataLoader
+from torchmetrics import MetricCollection
+from torchvision import datasets, transforms
+from torchmetrics.classification import Accuracy, Precision, Recall, F1Score
+
+from PIL import Image
+from beartype import beartype
+
+# %% Гиперпараметры
+num_epoch = 8
+batch_size = 64
+learning_rate = 1e-3
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+# %% Трансформация данных
+
+# Resize(size) — изменяет размер изображения.
+# Grayscale() — преобразование изображения в оттенки серого.
+# ToTensor() — преобразует изображение PIL или NumPy-массив в PyTorch-тензор.
+# Normalize(mean, std) — нормализация тензора изображения.
+# GaussianBlur(kernel_size) — размытие по Гауссу.
+
+
+image_transforms = transforms.Compose([
+    transforms.ToTensor()
+])
+
+# %% Загрузка набора данных
+
+train_dataset = datasets.MNIST(
+    root="./MNIST",
+    train=True,
+    download=False, # Если скачен, то убрать
+    transform=image_transforms
+)
+
+test_dataset = datasets.MNIST(
+    root="./MNIST",
+    train=False,
+    download=False, # Если скачен, то убрать
+    transform=image_transforms    
+)
+
+train_loader = DataLoader(dataset=train_dataset,
+                          batch_size=batch_size,
+                          shuffle=True)
+
+test_loader = DataLoader(dataset=test_dataset,
+                         batch_size=batch_size,
+                         shuffle=False)
+# %% Структура модели
+
+class FFN(nn.Module):
+    
+    @beartype
+    def __init__(self,
+                 input_size: int,
+                 output_size: int):
+        super().__init__()
+        
+        self.input_layer = nn.Linear(input_size, 32*32)
+        self.hidden_layer_1 = nn.Linear(32*32, 32*32)
+        self.hidden_layer_2 = nn.Linear(32*32, 32*32)
+        self.output_layer = nn.Linear(32*32, output_size)
+    
+    @beartype
+    def forward(self, image: torch.Tensor) -> torch.Tensor:
+        x = image.flatten(start_dim=1)  # batch_size x 28*28
+        x = torch.relu(self.input_layer(x))
+        x = torch.relu(self.hidden_layer_1(x))
+        x = torch.relu(self.hidden_layer_2(x))
+        x = self.output_layer(x)
+        
+        return x # batch_size x 10
+
+model = FFN(input_size=28*28,
+            output_size=10)
+model.to(device)
+
+# %% Метрики
+
+train_metrics = MetricCollection({
+    "accuracy": Accuracy(task="multiclass", num_classes=10),
+    "precision": Precision(task="multiclass", num_classes=10, average="macro"),    
+    "recall": Recall(task="multiclass", num_classes=10, average="macro"),
+    "f1": F1Score(task="multiclass", num_classes=10, average="macro")    
+}).to(device)
+
+test_metrics = MetricCollection({
+    "accuracy": Accuracy(task="multiclass", num_classes=10),
+    "precision": Precision(task="multiclass", num_classes=10, average="macro"),    
+    "recall": Recall(task="multiclass", num_classes=10, average="macro"),
+    "f1": F1Score(task="multiclass", num_classes=10, average="macro")    
+}).to(device)
+
+# %% Оптимизаторы
+
+criterion = nn.CrossEntropyLoss()
+optim = torch.optim.Adam(model.parameters(), lr=learning_rate)
+# %% Тренировочный цикл
+
+for epoch in tqdm.tqdm(range(num_epoch)):
+    model.train()
+    running_loss = 0
+    
+    for data, target in train_loader:
+        data, target = data.to(device), target.to(device)
+        
+        optim.zero_grad()
+        outputs = model(data)
+        loss = criterion(outputs, target)
+        loss.backward()
+        running_loss += loss.item()
+        optim.step()
+        
+        preds = outputs.argmax(dim=1)
+        train_metrics.update(preds, target)
+    
+    results = train_metrics.compute()        
+    print(f"Epoch: {epoch+1}")
+    print("=== TRAIN ===")
+    print(f"  Loss: {(running_loss/len(train_loader)):.4f}")
+    print(f"  Accuracy:  {results['accuracy']:.4f}")
+    print(f"  Precision: {results['precision']:.4f}")
+    print(f"  Recall:    {results['recall']:.4f}")
+    print(f"  F1:        {results['f1']:.4f}")
+    print()
+    
+    train_metrics.reset()
+    
+    model.eval()
+    with torch.no_grad():
+        test_loss = 0
+        for data, target in test_loader:
+            data, target = data.to(device), target.to(device)
+            
+            outputs = model(data)
+            loss = criterion(outputs, target)
+            test_loss += loss.item()
+            preds = outputs.argmax(dim=1)
+            test_metrics.update(preds, target)
+        
+        results = test_metrics.compute()        
+        
+        print("=== TEST ===")
+        print(f"  Loss: {(test_loss/len(train_loader)):.4f}")
+        print(f"  Accuracy:  {results['accuracy']:.4f}")
+        print(f"  Precision: {results['precision']:.4f}")
+        print(f"  Recall:    {results['recall']:.4f}")
+        print(f"  F1:        {results['f1']:.4f}")
+        print()
+        
+    test_metrics.reset()
+        
+    
+torch.save(model.state_dict(), "./models/mnist_model.pth")
+# %% Инференс модели
+
+image_transforms = transforms.Compose([
+    transforms.Grayscale(),
+    transforms.Resize((28,28)),
+    transforms.ToTensor()    
+])
+
+model.eval()
+model.load_state_dict(torch.load("./models/mnist_model.pth"))
+
+def predict(image_path):
+    image = Image.open(image_path)
+    image = image_transforms(image)
+    image = image.unsqueeze(0)
+    image = image.to(device)
+    
+    with torch.no_grad():
+        output = model(image)
+        
+        pred = output.argmax(dim=1).item()
+        
+    return pred
+
+print(f"Predicted label: {predict('./test_images/mnist/3.png')}")
+print(f"Predicted label: {predict('./test_images/mnist/5.png')}")
+print(f"Predicted label: {predict('./test_images/mnist/6.png')}")