Spaces:
Running on Zero
Running on Zero
| { | |
| "slug": "adaptive-avgpool2d", | |
| "title": "手撕全局自适应平均池化 (AdaptiveAvgPool2d)", | |
| "difficulty": "Easy", | |
| "category": "DL", | |
| "tags": [ | |
| "自适应池化", | |
| "CNN" | |
| ], | |
| "description_md": "### 题目描述\n将任意尺寸的特征图自适应池化到指定的输出高宽 $(H_{out}, W_{out})$。\n\n### 函数签名\n```python\ndef adaptive_avgpool2d(\n x: torch.Tensor,\n output_size: Tuple[int, int],\n) -> torch.Tensor:\n```\n\n### 示例\n```python\n# 输入\nx = [[[[1.0, 2.0], [3.0, 4.0]], [[-1.0, 0.5], [1.5, -0.5]]], [[[2.5, 0.0], [-1.5, 3.5]], [[1.0, 2.0], [3.0, 4.0]]]]\noutput_size = (1, 1)\n\n# 输出 (Tensor, shape=[2, 2, 1, 1], dtype=float32)\n[[[[2.5]], [[0.125]]], [[[1.125]], [[2.5]]]]\n```\n\n### 参数规格与约束\n- **`x`**：`torch.Tensor`，测试规格 `shape=[2, 4, 14, 14], dtype=float32`\n- **`output_size`**：`Tuple[int, int]`\n- **返回值**：`torch.Tensor`，输出规格 `shape=[2, 4, 1, 1], dtype=float32`\n- **精度容差**：`atol=0.0001, rtol=0.0001`", | |
| "entry_point": "adaptive_avgpool2d", | |
| "starter_code": "import torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nimport math\nfrom typing import Optional, List, Tuple\n\ndef adaptive_avgpool2d(\n x: torch.Tensor,\n output_size: Tuple[int, int],\n) -> torch.Tensor:\n \"\"\"\n 手撕全局自适应平均池化 (AdaptiveAvgPool2d)\n \"\"\"\n # 请在此编写代码\n pass\n", | |
| "reference_code": "import torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nimport math\nfrom typing import Optional, List, Tuple\n\ndef adaptive_avgpool2d(x: torch.Tensor, output_size: Tuple[int, int]) -> torch.Tensor:\n return F.adaptive_avg_pool2d(x, output_size)\n", | |
| "forbidden_modules": [ | |
| "os", | |
| "sys", | |
| "subprocess", | |
| "shutil", | |
| "socket", | |
| "requests" | |
| ], | |
| "forbidden_calls": [], | |
| "time_limit": 3.0, | |
| "memory_limit_mb": 512, | |
| "atol": 0.0001, | |
| "rtol": 0.0001, | |
| "test_code": "import torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nimport math\n\ncases = [\n ('自适应池化', torch.randn(2, 4, 14, 14), (1, 1)),\n ('多批次扩展泛化验证', torch.randn(4, 4, 14, 14), (1, 1)),\n ('多维批次与参数泛化验证', torch.randn(2, 4, 14, 14), (1, 1)),\n ('边界极值与稳定性验证', torch.randn(2, 4, 14, 14), (1, 1)),\n]\nfor idx, (name, *args) in enumerate(cases, 1):\n t0 = time.time()\n try:\n exp = ref_obj(*args)\n act = user_obj(*args)\n match, err, max_diff = compare_outputs(act, exp, atol=atol, rtol=rtol)\n passed, msg = match, err\n except Exception as e:\n passed, msg = False, f\"Runtime Exception: {str(e)}\"\n act, exp, max_diff = None, None, None\n cases_results.append({\n \"case_id\": idx, \"case_name\": name, \"passed\": passed,\n \"time_ms\": round((time.time() - t0) * 1000, 2), \"message\": msg,\n \"user_output_summary\": summarize_obj(act) if act is not None else None,\n \"expected_summary\": summarize_obj(exp) if exp is not None else None,\n \"max_diff\": max_diff,\n })\n", | |
| "is_featured": false, | |
| "editorial_md": "#### 💡 核心算法原理与数学推导\n**手撕全局自适应平均池化 (AdaptiveAvgPool2d)**（`adaptive_avgpool2d`）聚焦于 **DL** 中的关键计算逻辑：\n\n将任意尺寸的特征图自适应池化到指定的输出高宽 $(H_{out}, W_{out})$。\n\n在实现中，函数接收参数 `x` (`torch.Tensor`)、`output_size` (`Tuple[int, int]`)，目标是计算出类型与形状严格符合 `torch.Tensor` 契约的结果。\n\n#### 📐 输入输出张量变换与步骤拆解\n- **输入基准规格**：`x` 规格为 `torch.Tensor(shape=[2, 4, 14, 14], dtype=float32)`，`output_size` 规格为 `Tuple` (`int` (示例值: `1`), `int` (示例值: `1`))。\n- **期望输出规格**：`torch.Tensor(shape=[2, 4, 1, 1], dtype=float32)`。\n\n**具体推导与计算步骤如下：**\n1. 返回最终结果：输出 `F.adaptive_avg_pool2d(x, output_size)`，严格对齐函数签名要求的返回类型与维度。\n\n#### 📊 复杂度分析\n- **时间复杂度**：$O(B \\cdot C \\cdot H_{out} \\cdot W_{out} \\cdot K_h \\cdot K_w)$（或对应动态规划网格规模 $O(T_1 \\cdot T_2)$），取决于滑动窗口或递推网格遍历次数。\n- **空间复杂度**：$O(B \\cdot C \\cdot H_{out} \\cdot W_{out})$，用于存储输出特征图或递推状态表。\n\n#### ⚠️ 关键实现细节与易错点\n- **输入输出类型与形状契约**：请严格按照函数签名返回对应类型（`torch.Tensor`、`Tuple` 或标量数值），避免遗漏批次维度或改变张量的数据类型（`dtype`）。\n- **边界输入鲁棒性**：注意单样本（`Batch=1`）、单步序列（`Seq=1`）、零值输入或极端超参数下的计算行为，不要在代码中硬编码特定测试用例的维度常数。\n- **纯向量化算子优先**：尽量使用 PyTorch 原生向量化张量算子替代低效的 Python 逐元素循环，以兼顾代码简洁性与执行性能。\n\n#### 💻 官方标准参考实现\n```python\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nimport math\nfrom typing import Optional, List, Tuple\n\ndef adaptive_avgpool2d(x: torch.Tensor, output_size: Tuple[int, int]) -> torch.Tensor:\n return F.adaptive_avg_pool2d(x, output_size)\n```", | |
| "difficulty_breakdown": { | |
| "concept": 1, | |
| "implementation": 1, | |
| "math_derivation": 2, | |
| "boundary_complexity": 1 | |
| }, | |
| "test_cases": [ | |
| { | |
| "name": "自适应池化", | |
| "input_desc": "真实参数契约用例 (x, output_size)", | |
| "params": [ | |
| { | |
| "name": "x", | |
| "type": "torch.Tensor", | |
| "value": "torch.tensor([[[[1.0, 2.0], [3.0, 4.0]], [[-1.0, 0.5], [1.5, -0.5]]], [[[2.5, 0.0], [-1.5, 3.5]], [[1.0, 2.0], [3.0, 4.0]]]], dtype=torch.float32)" | |
| }, | |
| { | |
| "name": "output_size", | |
| "type": "Tuple[int, int]", | |
| "value": "(1, 1)" | |
| } | |
| ], | |
| "raw_inputs": { | |
| "x": "torch.tensor([[[[1.0, 2.0], [3.0, 4.0]], [[-1.0, 0.5], [1.5, -0.5]]], [[[2.5, 0.0], [-1.5, 3.5]], [[1.0, 2.0], [3.0, 4.0]]]], dtype=torch.float32)", | |
| "output_size": "(1, 1)" | |
| } | |
| }, | |
| { | |
| "name": "多批次扩展泛化验证", | |
| "input_desc": "真实参数契约用例 (x, output_size)", | |
| "params": [ | |
| { | |
| "name": "x", | |
| "type": "torch.Tensor", | |
| "value": "torch.randn(4, 4, 14, 14)" | |
| }, | |
| { | |
| "name": "output_size", | |
| "type": "Tuple[int, int]", | |
| "value": "(1, 1)" | |
| } | |
| ], | |
| "raw_inputs": { | |
| "x": "torch.randn(4, 4, 14, 14)", | |
| "output_size": "(1, 1)" | |
| } | |
| }, | |
| { | |
| "name": "多维批次与参数泛化验证", | |
| "input_desc": "真实参数契约用例 (x, output_size)", | |
| "params": [ | |
| { | |
| "name": "x", | |
| "type": "torch.Tensor", | |
| "value": "torch.randn(2, 4, 14, 14)" | |
| }, | |
| { | |
| "name": "output_size", | |
| "type": "Tuple[int, int]", | |
| "value": "(1, 1)" | |
| } | |
| ], | |
| "raw_inputs": { | |
| "x": "torch.randn(2, 4, 14, 14)", | |
| "output_size": "(1, 1)" | |
| } | |
| } | |
| ] | |
| } |