40 lines
1.2 KiB
Plaintext
40 lines
1.2 KiB
Plaintext
{
|
|
"cells": [
|
|
{
|
|
"cell_type": "markdown",
|
|
"id": "88dd3320",
|
|
"metadata": {
|
|
"origin_pos": 0
|
|
},
|
|
"source": [
|
|
"# 计算性能\n",
|
|
":label:`chap_performance`\n",
|
|
"\n",
|
|
"在深度学习中，数据集和模型通常都很大，导致计算量也会很大。\n",
|
|
"因此，计算的性能非常重要。\n",
|
|
"本章将集中讨论影响计算性能的主要因素：命令式编程、符号编程、\n",
|
|
"异步计算、自动并行和多GPU计算。\n",
|
|
"通过学习本章，对于前几章中实现的那些模型，可以进一步提高它们的计算性能。\n",
|
|
"例如，我们可以在不影响准确性的前提下，大大减少训练时间。\n",
|
|
"\n",
|
|
":begin_tab:toc\n",
|
|
" - [hybridize](hybridize.ipynb)\n",
|
|
" - [async-computation](async-computation.ipynb)\n",
|
|
" - [auto-parallelism](auto-parallelism.ipynb)\n",
|
|
" - [hardware](hardware.ipynb)\n",
|
|
" - [multiple-gpus](multiple-gpus.ipynb)\n",
|
|
" - [multiple-gpus-concise](multiple-gpus-concise.ipynb)\n",
|
|
" - [parameterserver](parameterserver.ipynb)\n",
|
|
":end_tab:\n"
|
|
]
|
|
}
|
|
],
|
|
"metadata": {
|
|
"language_info": {
|
|
"name": "python"
|
|
},
|
|
"required_libs": []
|
|
},
|
|
"nbformat": 4,
|
|
"nbformat_minor": 5
|
|
} |