未验证 提交 9807b754 编写于 作者: C cinderellaTiger 提交者: GitHub

Update introduction_cn.ipynb (#5720)

上级 56ac84f4
...@@ -7,11 +7,11 @@ ...@@ -7,11 +7,11 @@
"source": [ "source": [
"## 1. PP-HelixFold模型简介\n", "## 1. PP-HelixFold模型简介\n",
"\n", "\n",
"AlphaFold2是一款高精度的蛋白质结构预测模型。PP-HelixFold基于PaddlePaddle框架在GPU和DCU上完整复现了AlphaFold2的训练和推理流程,并进一步提升模型性能与精度。通过与原版AlphaFold2模型和哥伦比亚大学Mohammed AlQuraishi教授团队基于PyTorch复现的OpenFold模型的性能对比测试显示,PP-HelixFold将训练耗时从约11天减少到7.5天。在性能大幅度提升的同时,PP-HelixFold从头端到端完整训练可以达到AlphaFold2论文媲美的精度。\n", "AlphaFold2是一款高精度的蛋白质结构预测模型。PP-HelixFold基于PaddlePaddle框架在GPU和DCU上完整复现了AlphaFold2的训练和推理流程,并进一步提升模型性能与精度。通过与原版AlphaFold2模型和哥伦比亚大学Mohammed AlQuraishi教授团队基于PyTorch复现的OpenFold模型的性能对比测试显示,PP-HelixFold将训练耗时从约11天减少到5.12天,在使用混合并行时只需要2.89 天。在性能大幅度提升的同时,PP-HelixFold从头端到端完整训练可以达到AlphaFold2论文媲美的精度。\n",
"\n", "\n",
"<p align=\"center\">\n", "<p align=\"center\">\n",
"<img src=\"https://github.com/PaddlePaddle/PaddleHelix/blob/dev/.github/HelixFold_computational_performance.png?raw=true\" align=\"middle\" height=\"50%\" width=\"50%\" />\n", "<img src=\"https://github.com/PaddlePaddle/PaddleHelix/blob/dev/.github/HelixFold_computational_perf.png?raw=true\" align=\"middle\" height=\"50%\" width=\"50%\" />\n",
"<img src=\"https://github.com/PaddlePaddle/PaddleHelix/blob/dev/.github/HelixFold_accuracy.png?raw=true\" align=\"middle\" height=\"60%\" width=\"60%\" />\n", "<img src=\"https://github.com/PaddlePaddle/PaddleHelix/blob/dev/.github/HelixFold_infer_accuracy.png?raw=true\" align=\"middle\" height=\"60%\" width=\"60%\" />\n",
"</p>\n", "</p>\n",
"\n", "\n",
"\n", "\n",
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册