# 项目文件：demos/loader_batches.py
# 五条样本按每批两条读取，最后一批保留一条；固定生成器后核对打乱顺序。
import torch
from torch.utils.data import DataLoader, TensorDataset

x = torch.arange(5, dtype=torch.float32).unsqueeze(1)
y = 2 * x + 1
dataset = TensorDataset(x, y)
for features, target in DataLoader(dataset, batch_size=2):
    print("本批", features[:, 0].tolist(), tuple(features.shape), tuple(target.shape))


def order():
    generator = torch.Generator().manual_seed(7)
    loader = DataLoader(dataset, batch_size=2, shuffle=True, generator=generator)
    return [number for features, _ in loader for number in features[:, 0].tolist()]


print("固定生成器顺序相同", order() == order())

# 项目文件：training_project/src/readingtorch/data.py
import csv
from dataclasses import dataclass
import hashlib
from pathlib import Path

import numpy as np
import torch
from torch.utils.data import DataLoader, Dataset, Subset


class ArrayDataset(Dataset):
    # 一次索引返回一条输入和对应目标，批量组合交给 DataLoader。
    def __init__(self, x, y):
        self.x, self.y = x, y

    def __len__(self):
        return len(self.x)

    def __getitem__(self, index):
        return self.x[index], self.y[index]


def read_raw(task, path):
    if task == "regression":
        with path.open(encoding="utf-8", newline="") as stream:
            rows = list(csv.DictReader(stream))
        x = torch.tensor([[float(r["minutes"]), float(r["sessions"])] for r in rows])
        y = torch.tensor([[float(r["pages"])] for r in rows])
    elif task == "images":
        with np.load(path, allow_pickle=False) as archive:
            x = torch.from_numpy(archive["images"].astype(np.float32)).unsqueeze(1)
            y = torch.from_numpy(archive["labels"].astype(np.int64))
    else:
        raise ValueError("请选择 regression 或 images")
    if not torch.isfinite(x).all() or not torch.isfinite(y).all():
        raise ValueError("数据中出现了 NaN 或无穷值")
    return x.float(), y