-
Notifications
You must be signed in to change notification settings - Fork 73
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
xusuyong add 2023.10.10~2023.10.24 WeeklyReport
- Loading branch information
Showing
2 changed files
with
30 additions
and
0 deletions.
There are no files selected for viewing
30 changes: 30 additions & 0 deletions
30
WeeklyReports/13_xusuyong/[WeeklyReport]2023.10.10~2023.10.24.md
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,30 @@ | ||
### 姓名 | ||
|
||
徐苏勇 | ||
|
||
Github ID:[xusuyong](https://github.com/xusuyong) | ||
|
||
### 实习项目 | ||
|
||
[科学计算领域拓展专项](https://github.com/PaddlePaddle/community/blob/master/hackathon/hackathon_5th/%E3%80%90PaddlePaddle%20Hackathon%205th%E3%80%91%E9%A3%9E%E6%A1%A8%E6%8A%A4%E8%88%AA%E8%AE%A1%E5%88%92%E9%9B%86%E8%AE%AD%E8%90%A5%E9%A1%B9%E7%9B%AE%E5%90%88%E9%9B%86.md#%E9%A1%B9%E7%9B%AE%E5%8D%81%E4%B8%89%E7%A7%91%E5%AD%A6%E8%AE%A1%E7%AE%97%E9%A2%86%E5%9F%9F%E6%8B%93%E5%B1%95%E4%B8%93%E9%A1%B9) | ||
|
||
### 本周工作 | ||
|
||
1. **改造PaddleScience下的案例laplace2d的代码为Hydra的形式** | ||
- 理解并跑通了DeepONet、laplace2d等案例,熟悉了PaddleScience用PINN和神经算子的算法解偏微分方程的基本逻辑 | ||
- 改造laplace2d案例的代码的代码,使之能使用[Hydra](https://hydra.cc/)库,方便实验管理和解析。PR地址:[modify laplace2d to hydra style](https://github.com/PaddlePaddle/PaddleScience/pull/575) | ||
2. **学习理解用DDPM算法从低分辨数据重构高分辨率数据** | ||
- 学习理解DDPM算法,复现文献 **A Physics-informed Diffusion Model for High-fidelity Flow Field Reconstruction** (<a href="https://www.sciencedirect.com/science/article/pii/S0021999123000670">Journal of Computational Physics</a> | <a href="https://arxiv.org/abs/2211.14680">arXiv</a>),理解作者是如何将物理知识融入DDPM算法,模型预测的结果: | ||
![](pred.png) | ||
- 学习使用SU2 | ||
3. **问题疑惑与解答** | ||
- 问题:原始的DDPM/DDIM采样算法是从输入高斯噪声开始,生成的是随机的图片难以控制,如何让它生成我们想要的高分辨率数据? | ||
|
||
答:为了解决这个问题,需要一个有指导的数据生成程序,其中使用低精度CFD数据作为生成高精度CFD数据的条件。反向扩散过程的马尔可夫性质意味着生成x0的过程不必从xT开始,而是可以从任何时间步长t∈{1,…,t}开始,前提是xT可用。该属性允许用户在后向扩散过程的特定时间步长选择中间样本,并将其发送到马尔可夫链的剩余部分以获得x0。 | ||
### 下周工作 | ||
|
||
1. 学习物理信息扩散模型,理解它是如何将物理信息加入DDPM算法的。 | ||
2. 调研SU2与DDPM的结合使用方式 | ||
|
||
### 导师点评 | ||
|
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.