{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 1 章节介绍\n",
    "\n",
    "本教程《Sana-Video 推理优化实践》以 `Sana-Video` 为例，展示如何在昇腾 NPU 上完成模型跑通、Profiling 分析，并在整网中接入 `RMSNorm` 融合算子验证性能收益。\n",
    "\n",
    "---\n",
    "\n",
    "## 学习目标\n",
    "- 完成 `Sana-Video` Baseline 跑通。\n",
    "- 使用 `torch_npu.profiler` 采集 Baseline 性能数据。\n",
    "- 在 `Sana-Video` 中接入 `torch_npu.npu_rms_norm` 并验证整网收益。\n",
    "- 对比优化前后的整网时延变化。\n",
    "\n",
    "## 章节安排\n",
    "- [1 章节介绍](./01_chapter_intro.ipynb)\n",
    "- [2 Baseline 跑通](./02_baseline_run.ipynb)\n",
    "- [3 Profiling 分析](./03_profiling_analysis.ipynb)\n",
    "- [4 RMSNorm 融合接入与收益验证](./04_rmsnorm_fusion_optimization.ipynb)\n",
    "\n",
    "## 环境说明\n",
    "- 本教程依赖上游 `Sana` 仓库代码，并在运行目录中拉取固定 commit。\n",
    "- 在线体验请直接在 GitCode Notebook 环境中执行；Notebook 默认复用环境中已预装的 `torch`、`torch_npu`、`torchvision` 与 `torchaudio`。本地运行前请先准备兼容版本的上述依赖，并配置 CANN 与 `torch_npu`。\n",
    "- 运行本教程时，主机内存需至少 16GB。Notebook 计算类型建议选择 NPU 910B、CPU 32GB，容器镜像建议选择 ubuntu22-cann8.5-py3.11-jupyter-notebook。\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}