MLP优化

[hongxiahao91]

Task Description

我们这边原子模拟的模型，graph transformer架构。现在有三个问题：1. 现在模型设计上有点问题，模型的势能面不够平滑，导致在做材料的结构优化的时候fail不好收敛。 2. 模型attention部分比较原始，导致memory特别大，模型训练特别吃显存。3. 模型架构上，transformer领域的各种先进的技术，没有尝试过，不知是否会有很好的机会在里面。

Baseline Repository Link (Must be Public)

TBD

Baseline reproduction (minimal)

TBD

Dataset (Must be Public)

TBD

Results and Evaluation Metrics

matbench discovery board

Preconditions (required)

• The baseline I provide comes from public code hosted on an open-source platform such as GitHub.
• The task I use/optimize is based on a public dataset.
• If my request involves private code/data, I will contact interndiscovery@pjlab.org.cn by email instead.

[Shiyang980713]

Hi, hongxiahao91
Thanks for submitting your task!

Very interesting topic! According to our task template, you still need to provide at least the baseline code repository and dataset. If the code/data is private, please send it to interndiscovery@pjlab.org.cn.

Looking forward to your update, and feel free to reach out if you have any questions! 😊

Best,
InternAgent Team

[Shiyang980713]

Hi @hongxiahao91 ,

The full codebase of InternAgent 1.5 has now been open-sourced. Feel free to use our code to try optimizing your project.

Thank you!

Best,
InternAgent Team