Skip to content

Commit b301103

Browse files
committed
feat(ppsci): support data_effient_nopt
1 parent 14cdc71 commit b301103

File tree

2 files changed

+6
-105
lines changed

2 files changed

+6
-105
lines changed

docs/zh/examples/data_efficient_nopt.md

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -163,6 +163,12 @@ examples/data_efficient_nopt/data_efficient_nopt.py
163163

164164
总而言之,这篇论文提出了一种创新且高效的神经算子学习框架,通过无监督预训练在大量廉价的无标签物理数据上学习通用表示,并通过情境学习在推理阶段利用少量相似案例来提升OOD泛化能力。这一框架显著降低了对昂贵模拟数据的需求,并提高了模型在复杂物理问题中的适应性和泛化性,为科学机器学习的数据高效发展开辟了新途径。
165165

166+
下方展示了部分实验结果:
167+
168+
| Model | Checkpoint | **$RMSE$** | **RMSE (normalized)$** | **R2** | **Slope** |
169+
| :---------: | :--------------------: | :--------: | :--------------------: | :----: | :-------: |
170+
| FNO_Possion | finetune_b01_m0_n8192 | 0.2586 | 0.1414 | 0.9765 | 0.9752 |
171+
166172
## 6. 参考资料
167173

168174
- [Data-Efficient Operator Learning via Unsupervised Pretraining and In-Context Learning](https://arxiv.org/abs/2402.15734)

examples/data_efficient_nopt/config/vmae_config_pretrain.yaml

Lines changed: 0 additions & 105 deletions
This file was deleted.

0 commit comments

Comments
 (0)