{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/duoan/TorchCode/blob/master/templates/16_cross_entropy.ipynb)\n",
    "\n",
    "# 🟢 Easy: Cross-Entropy Loss\n",
    "\n",
    "Implement **cross-entropy loss** from scratch.\n",
    "\n",
    "$$\\text{CE}(x, y) = -\\log\\frac{e^{x_y}}{\\sum_j e^{x_j}}$$\n",
    "\n",
    "### Signature\n",
    "```python\n",
    "def cross_entropy_loss(logits: Tensor, targets: Tensor) -> Tensor:\n",
    "    # logits: (B, C) float, targets: (B,) long indices\n",
    "    # Returns: scalar loss (mean over batch)\n",
    "```\n",
    "\n",
    "### Rules\n",
    "- Do NOT use `F.cross_entropy` or `nn.CrossEntropyLoss`\n",
    "- Must be numerically stable (use logsumexp trick)"
   ],
   "outputs": []
  },
  {
   "cell_type": "code",
   "metadata": {},
   "source": [
    "# Install torch-judge in Colab (no-op in JupyterLab/Docker)\n",
    "try:\n",
    "    import google.colab\n",
    "    get_ipython().run_line_magic('pip', 'install -q torch-judge')\n",
    "except ImportError:\n",
    "    pass\n"
   ],
   "outputs": [],
   "execution_count": null
  },
  {
   "cell_type": "code",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch"
   ],
   "execution_count": null
  },
  {
   "cell_type": "code",
   "metadata": {},
   "outputs": [],
   "source": [
    "# ✏️ YOUR IMPLEMENTATION HERE\n",
    "\n",
    "def cross_entropy_loss(logits, targets):\n",
    "    pass  # log_probs = logits - logsumexp(...)"
   ],
   "execution_count": null
  },
  {
   "cell_type": "code",
   "metadata": {},
   "outputs": [],
   "source": [
    "# 🧪 Debug\n",
    "logits = torch.randn(4, 10)\n",
    "targets = torch.randint(0, 10, (4,))\n",
    "print('Loss:', cross_entropy_loss(logits, targets))\n",
    "print('Ref: ', torch.nn.functional.cross_entropy(logits, targets))"
   ],
   "execution_count": null
  },
  {
   "cell_type": "code",
   "metadata": {},
   "outputs": [],
   "source": [
    "# ✅ SUBMIT\n",
    "from torch_judge import check\n",
    "check('cross_entropy')"
   ],
   "execution_count": null
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "name": "python",
   "version": "3.11.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}