import unittest
import torch
from torch.utils.data import TensorDataset, DataLoader, RandomSampler
import importlib
import io
import sys
import pandas as pd


def _dynamic(test, cond, success, fail):
    if cond:
        test._testMethodName = success
        test.assertTrue(True)
    else:
        test._testMethodName = fail
        test.fail(fail)


class TestUserCode(unittest.TestCase):

    @classmethod
    def setUpClass(cls):
        cls.df = pd.read_csv(
            "https://content-media-cdn.codefinity.com/courses/1dd2b0f6-6ec0-40e6-a570-ed0ac2209666/section_2/iris.csv"
        )
        cls.features = torch.tensor(cls.df.drop(columns="species").values, dtype=torch.float32)
        cls.targets = torch.tensor(cls.df["species"].values, dtype=torch.long)

    # -------- Dataset exists --------
    def test_dataset_exists(self):
        import user_code
        cond = hasattr(user_code, "iris_dataset")
        _dynamic(
            self,
            cond,
            "`iris_dataset` is successfully created.",
            "`iris_dataset` is missing. Create it using `TensorDataset(features_tensor, target_tensor)`."
        )

    # -------- Dataset is TensorDataset --------
    def test_dataset_is_tensor_dataset(self):
        import user_code
        cond = isinstance(user_code.iris_dataset, TensorDataset)
        _dynamic(
            self,
            cond,
            "`iris_dataset` is a valid TensorDataset object.",
            "`iris_dataset` must be created using `TensorDataset(features_tensor, target_tensor)`."
        )

    # -------- Dataset content correctness --------
    def test_dataset_tensors_correct(self):
        import user_code
        ds = user_code.iris_dataset

        try:
            feats_ok = torch.allclose(ds.tensors[0], self.features)
            targets_ok = torch.allclose(ds.tensors[1], self.targets)
            cond = feats_ok and targets_ok
        except Exception:
            cond = False

        _dynamic(
            self,
            cond,
            "`iris_dataset` correctly stores the feature and target tensors.",
            "`iris_dataset` must contain `(features_tensor, target_tensor)` in that order."
        )

    # -------- DataLoader exists --------
    def test_dataloader_exists(self):
        import user_code
        cond = hasattr(user_code, "iris_loader")
        _dynamic(
            self,
            cond,
            "`iris_loader` is successfully created.",
            "`iris_loader` is missing. Wrap your dataset using `DataLoader(...)`."
        )

    # -------- DataLoader parameters --------
    def test_dataloader_params(self):
        import user_code
        loader = user_code.iris_loader

        cond = (
            isinstance(loader, DataLoader)
            and loader.batch_size == 32
            and isinstance(loader.sampler, RandomSampler)
        )

        _dynamic(
            self,
            cond,
            "`iris_loader` uses batch_size=32 and shuffles the data (RandomSampler).",
            "`iris_loader` must be created with `batch_size=32` and `shuffle=True` so that it uses a RandomSampler."
        )

    # -------- First batch correctness --------
    def test_first_batch_shape(self):
        import user_code

        loader = user_code.iris_loader
        batch_features, batch_targets = next(iter(loader))

        cond = (
            batch_features.ndim == 2
            and batch_features.shape[1] == 4
            and batch_targets.ndim == 1
        )

        _dynamic(
            self,
            cond,
            "Batches from `iris_loader` have the expected shapes.",
            "Batch shapes are incorrect. Features must be `[batch_size, 4]` and targets must be `[batch_size]`."
        )

    # -------- Output printing --------
    def test_output_printed(self):
        import user_code

        buf = io.StringIO()
        sys.stdout = buf
        importlib.reload(user_code)
        sys.stdout = sys.__stdout__

        cond = len(buf.getvalue().strip()) > 0

        _dynamic(
            self,
            cond,
            "The DataLoader output is printed successfully.",
            "No output detected. Ensure you print the first batch from the DataLoader."
        )


if __name__ == "__main__":
    unittest.main()


test_code.py

Passaggio dalle librerie di alto livello alla flessibilità di PyTorch. Padronanza dei componenti fondamentali del deep learning lavorando con i tensori, la struttura dati principale di PyTorch. Esplorazione della differenziazione automatica con i gradienti, implementazione della retropropagazione e costruzione di reti neurali complete da zero. Gestione di dataset reali e dell'intero ciclo di addestramento, dall'inizializzazione dei pesi alla valutazione del modello.

Sfida: Preparazione del Dataset Iris

Soluzione

Sfida: Preparazione del Dataset Iris

Soluzione