import unittest
import torch
import torch.nn as nn
import torch.optim as optim
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/hours_scores.csv"
        )
        cls.X = torch.tensor(cls.df["Hours Studied"].values, dtype=torch.float32).reshape(-1, 1)
        cls.Y = torch.tensor(cls.df["Test Score"].values, dtype=torch.float32).reshape(-1, 1)

    # ---------- INPUT TENSORS ----------
    def test_X_tensor_correct(self):
        import user_code
        cond = hasattr(user_code, "X_tensor")
        if cond:
            cond = torch.allclose(user_code.X_tensor, self.X) and user_code.X_tensor.ndim == 2
        _dynamic(
            self,
            cond,
            "`X_tensor` correctly contains reshaped hour values as a 2D tensor.",
            "`X_tensor` is incorrect. Ensure you convert X to a float tensor and reshape to [-1, 1]."
        )

    def test_Y_tensor_correct(self):
        import user_code
        cond = hasattr(user_code, "Y_tensor")
        if cond:
            cond = torch.allclose(user_code.Y_tensor, self.Y) and user_code.Y_tensor.ndim == 2
        _dynamic(
            self,
            cond,
            "`Y_tensor` correctly contains reshaped score values as a 2D tensor.",
            "`Y_tensor` is incorrect. Ensure you convert Y to a float tensor and reshape to [-1, 1]."
        )

    # ---------- MODEL ----------
    def test_model_is_linear(self):
        import user_code

        cond = isinstance(user_code.model, nn.Linear) and \
               user_code.model.in_features == 1 and user_code.model.out_features == 1

        _dynamic(
            self,
            cond,
            "The model is correctly defined as `nn.Linear(1, 1)`.",
            "The model must be `nn.Linear(1, 1)`."
        )

    # ---------- LOSS FUNCTION ----------
    def test_loss_function(self):
        import user_code

        cond = isinstance(user_code.criterion, nn.MSELoss)

        _dynamic(
            self,
            cond,
            "Loss function correctly set to MSE.",
            "Loss function must be `nn.MSELoss()`."
        )

    # ---------- OPTIMIZER ----------
    def test_optimizer(self):
        import user_code
        cond = isinstance(user_code.optimizer, optim.SGD)
        if cond:
            lr_correct = user_code.optimizer.param_groups[0]["lr"] == 0.01
            cond = cond and lr_correct

        _dynamic(
            self,
            cond,
            "Optimizer correctly set to SGD with learning rate 0.01.",
            "Optimizer must be `optim.SGD(model.parameters(), lr=0.01)`."
        )

    # ---------- TRAINING STEPS ----------
    def test_predictions_exist(self):
        import user_code
        cond = hasattr(user_code, "predictions")
        _dynamic(
            self,
            cond,
            "Predictions computed successfully during training.",
            "Predictions missing. Compute them using `model(X_tensor)`."
        )

    def test_loss_exists(self):
        import user_code
        cond = hasattr(user_code, "loss")
        _dynamic(
            self,
            cond,
            "Loss computed successfully during training.",
            "Loss missing. Compute it using `criterion(predictions, Y_tensor)`."
        )

    def test_weights_and_bias_extracted(self):
        import user_code

        cond = hasattr(user_code, "weights") and hasattr(user_code, "bias")
        if cond:
            cond = isinstance(user_code.weights, torch.Tensor) and isinstance(user_code.bias, torch.Tensor)

        _dynamic(
            self,
            cond,
            "Model parameters (weights and bias) successfully extracted.",
            "You must extract parameters using `model.weight.data` and `model.bias.data`."
        )

    # ---------- PRINT OUTPUT ----------
    def test_print_output(self):
        import user_code
        buffer = io.StringIO()
        sys.stdout = buffer

        importlib.reload(user_code)
        sys.stdout = sys.__stdout__

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

        _dynamic(
            self,
            cond,
            "Training logs and model parameters printed successfully.",
            "No output detected. Make sure to print loss updates and final parameters."
        )


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


test_code.py

Apprendi i concetti fondamentali e avanzati necessari per lavorare in modo efficiente con PyTorch. Ottieni una solida comprensione dei tensori, inclusa la creazione, le operazioni e il rimodellamento. Esplora gli elementi essenziali di gradienti, backpropagation e regressione lineare prima di passare alla gestione dei dataset. Padroneggia le competenze necessarie per costruire, addestrare e valutare reti neurali.

Esplora i fondamenti di PyTorch, concentrandoti sui tensori, la struttura dati principale utilizzata per i calcoli. Verranno trattati la creazione dei tensori, l'inizializzazione casuale, le operazioni matematiche e la manipolazione delle forme.

Esplorazione dei concetti chiave per l'addestramento dei modelli in PyTorch, inclusi il calcolo dei gradienti e la retropropagazione multi-step. Approfondimento della regressione lineare come modello fondamentale di machine learning e introduzione alla gestione efficiente dei dataset.

Scopri come costruire, addestrare e valutare reti neurali utilizzando PyTorch. Imparerai a definire una semplice rete neurale feedforward, ottimizzare i suoi parametri tramite l'addestramento e valutare le sue prestazioni.

Sfida: Implementazione della Regressione Lineare

Soluzione

Awesome!

Sfida: Implementazione della Regressione Lineare

Soluzione