diff --git a/docs/mindchemistry/docs/source_en/images/high-alloy.png b/docs/mindchemistry/docs/source_en/images/high-alloy.png deleted file mode 100644 index ce2e0c7e1bac9ff4a34bc0475f797a6792eac6a1..0000000000000000000000000000000000000000 Binary files a/docs/mindchemistry/docs/source_en/images/high-alloy.png and /dev/null differ diff --git a/docs/mindchemistry/docs/source_en/images/mindchemistry_arch.png b/docs/mindchemistry/docs/source_en/images/mindchemistry_arch.png deleted file mode 100644 index fae4dfcc250fcdaf33b44a283779889c382cc45d..0000000000000000000000000000000000000000 Binary files a/docs/mindchemistry/docs/source_en/images/mindchemistry_arch.png and /dev/null differ diff --git a/docs/mindchemistry/docs/source_en/images/mindchemistry_archi.png b/docs/mindchemistry/docs/source_en/images/mindchemistry_archi.png new file mode 100644 index 0000000000000000000000000000000000000000..5faa90c9ab036f24ac56ed2f246497d59a762597 Binary files /dev/null and b/docs/mindchemistry/docs/source_en/images/mindchemistry_archi.png differ diff --git a/docs/mindchemistry/docs/source_en/index.rst b/docs/mindchemistry/docs/source_en/index.rst index fb4041459d13f27deaca5e3d2009182d6ca89577..f2e986c649839580a616155e44a534c79024029b 100644 --- a/docs/mindchemistry/docs/source_en/index.rst +++ b/docs/mindchemistry/docs/source_en/index.rst @@ -1,5 +1,5 @@ MindSpore Chemistry -=================== +===================== Introduction ------------ @@ -26,13 +26,14 @@ efficiency, and seek to facilitate an innovative paradigm of joint research between AI and chemistry, providing experts with novel perspectives and efficient tools. -.. figure:: ./images/mindchemistry_arch.png + +.. figure:: ./images/mindchemistry_archi.png :alt: MindSpore Chemistry Architecture Latest News ----------- - -- 🔥\ ``2024.07.30`` MindChemistry 0.1.0 is released. +- 🔥`2025.03.30` MindChemistry 0.2.0 has been released, featuring several powerful applications, including NequIP, Allegro, DeephE3nn, Matformer, and DiffCSP. +- 🔥`2024.07.30` MindChemistry 0.1.0 has been released. Features -------- @@ -40,30 +41,7 @@ Features Applications ~~~~~~~~~~~~ -- Material Generation - - - **Scenario**\ ：Inorganic chemistry - - **Dataset**\ ：High-entropy alloy dataset. The high-entropy alloy - dataset includes the chemical composition of known high-entropy - alloys and thermodynamic properties of the alloys. It provides - chemical composition information such as the metal element types - and corresponding percentages as well as thermodynamic properties - such as magnetostrictive effects and Curie temperatures. - - **Task**\ ：High-entropy alloy composition design. We integrate - Machine learning-enabled high-entropy alloy discovery[1] approach - for designing novel high-entropy alloys with low thermal expansion - coefficients(TEC) in active learning fashion. In the active - learning circle, candidates of high-enropy alloys are firstly - generated based on the AI model and the candidate components are - filtered based on the prediction model and the predicted thermal - expansion coefficient calculated by the thermodynamics. Finally, - the researchers need to determine the final high-entropy alloy - components based on experimental verification. - -.. figure:: ./images/high-alloy.png - :alt: MindSpore high alloy Architecture - -- **Property Prediction**\ ： +- **Force Prediction**\ ： - **Scenario**\ ：Organic chemistry - **Dataset**: Revised Molecular Dynamics 17(rMD17). rMD17 dataset @@ -72,7 +50,7 @@ Applications such as the atomic numbers and positions as well as molecular property information such as energies and forces. - **Task**\ ：Molecular energy prediction. We integrate the NequIP - model [2] and Allegro model [3], according to the position of each + model [1] and Allegro model [2], according to the position of each atom in the molecular system and structure description of the atomic number information construction diagram, and calculate the energy of the molecular system based on the equivariant @@ -81,47 +59,37 @@ Applications .. figure:: ./images/nequip.png :alt: MindSpore nequip Architecture -- **Electronic Structure Prediction**\ ： +- **DFT Prediction**\ ： - - **Scenario**: Materials + - **Scenario**: Materials Chemistry - **Dataset**: Bilayer graphene dataset. The dataset contains descriptive information such as atomic positions and atomic numbers, as well as property information such as Hamiltonian. - **Task**: Density Functional Theory Hamiltonian Prediction. We - integrate the DeephE3nn model [4], an equivariant neural network based + integrate the DeephE3nn model [3], an equivariant neural network based on E3, to predict a Hamiltonian by using the structure of atoms. -- **Prediction of crystalline material properties**: +- **Property Prediction**: - - **Scenario**: Materials + - **Scenario**: Materials Chemistry - **Dataset**: JARVIS-DFT 3D dataset. The dataset contains descriptive information such as atomic position and atomic number of crystal materials, as well as property information such as energy and force field. - **Task**: Prediction of crystalline material properties. We - integrate the Matformer model [5] based on graph neural networks and + integrate the Matformer model [4] based on graph neural networks and Transformer architectures, for predicting various properties of crystalline materials. -Modules -~~~~~~~ - -- **Equivariant Computing** - - - **Introduction**\ ：Symmetry is an essential property in science - domain. Equivarient neural network adopts intuitive representation - as input and computing equivariently with respect to spatial - rotation, shift and inversion. Adopting equivariant neural network - for modeling scientific scenarios results in higher representation - effectiveness for data and high efficiency for model training. - - **Functions**\ ：E(3) computing modules integrates basic modules - such as Irreps, Spherical Harmonics and Tensor Products. Based on - the basic modules, equivariant neural network layers such as - equivariant Activation, Linear and Convolution layers are provided - for constructing user customed equivariant neural networks. +- **Structure Generation**: -.. figure:: ./images/e3.png - :alt: MindSpore e3 Architecture + - **Scenario**: Materials Chemistry + - **Dataset**: + - Perov-5: A perovskite dataset in which each unit cell contains five fixed atoms, and the structures are relatively similar. + - Carbon-24: A carbon crystal dataset, where each crystal contains between 6 and 24 carbon atoms, with various different material structures. + - MP-20: A dataset collected from the MP database, featuring experimental structures with up to 20 atoms per unit cell. The materials and structures are highly diverse. + - MPTS-52: An advanced version of MP-20, expanding the number of atoms per unit cell to 52. The materials and structures are highly diverse. + - **Task**: Crystal material structure prediction. We integrated the DiffCSP model [5], which is based on a graph neural network and diffusion model architecture, to predict the crystal material structures given their composition. Installation ------------ @@ -194,10 +162,9 @@ Community Core Contributor ~~~~~~~~~~~~~~~~ -Thanks goes to these wonderful people 🧑‍🤝‍🧑: +Thanks goes to these wonderful people: -yufan, wangzidong, liuhongsheng, gongyue, gengchenhua, linghejing, -yanchaojie, suyun, wujian, caowenbin +wujian, wangyuheng, Lin Peijia, gengchenhua, caowenbin，Siyu Yang Contribution Guide ------------------ @@ -213,25 +180,15 @@ License References ---------- -[1] Rao Z, Tung P Y, Xie R, et al. Machine learning-enabled high-entropy -alloy discovery[J]. Science, 2022, 378(6615): 78-85. +[1] Batzner S, Musaelian A, Sun L, et al. E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials[J]. Nature communications, 2022, 13(1): 2453. -[2] Batzner S, Musaelian A, Sun L, et al. E(3)-equivariant graph neural -networks for data-efficient and accurate interatomic potentials[J]. -Nature communications, 2022, 13(1): 2453. +[2] Musaelian A, Batzner S, Johansson A, et al. Learning local equivariant representations for large-scale atomistic dynamics[J]. Nature communications, 2023, 14(1): 579. -[3] Musaelian A, Batzner S, Johansson A, et al. Learning local -equivariant representations for large-scale atomistic dynamics[J]. -Nature communications, 2023, 14(1): 579. +[3] Xiaoxun Gong, He Li, Nianlong Zou, et al. General framework for E(3)-equivariant neural network representation of density functional theory Hamiltonian[J]. Nature communications, 2023, 14: 2848. -[4] Xiaoxun Gong, He Li, Nianlong Zou, et al. General framework for -E(3)-equivariant neural network representation of density functional -theory Hamiltonian[J]. -Nature communications, 2023, 14: 2848. +[4] Keqiang Yan, Yi Liu, Yuchao Lin, Shuiwang ji, et al. Periodic Graph Transformers for Crystal Material Property Prediction[J]. arXiv:2209.11807v1 [cs.LG] 23 sep 2022. -[5] Keqiang Yan, Yi Liu, Yuchao Lin, Shuiwang ji, et al. Periodic -Graph Transformers for Crystal Material Property Prediction[J]. -arXiv:2209.11807v1 [cs.LG] 23 sep 2022. +[5] Jiao Rui and Huang Wenbing and Lin Peijia, et al. Crystal structure prediction by joint equivariant diffusion[J]. Advances in Neural Information Processing Systems, 2024, 36. .. toctree:: :maxdepth: 1 @@ -246,7 +203,7 @@ arXiv:2209.11807v1 [cs.LG] 23 sep 2022. :caption: User Guide :hidden: - user/molecular_generation + user/structure_generation user/molecular_prediction .. toctree:: diff --git a/docs/mindchemistry/docs/source_en/user/molecular_generation.md b/docs/mindchemistry/docs/source_en/user/molecular_generation.md deleted file mode 100644 index 69d8a7b22e29d3cb83ed742f42e3e3caff6c6d16..0000000000000000000000000000000000000000 --- a/docs/mindchemistry/docs/source_en/user/molecular_generation.md +++ /dev/null @@ -1,11 +0,0 @@ -# Molecular Generation - -[![View Source On Gitee](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/website-images/master/resource/_static/logo_source_en.svg)](https://gitee.com/mindspore/docs/blob/master/docs/mindchemistry/docs/source_zh_cn/user/molecular_generation.md) - -Molecular generation, using deep learning generation models to predict and generate components in the particle system. We have integrated a method based on active learning for high entropy alloy design, designing high entropy alloy components with extremely low thermal expansion coefficients. In the active learning process, first, candidate high entropy alloy components are generated based on AI models, and the candidate components are screened based on predictive models and thermodynamic calculations to predict the thermal expansion coefficient. Finally, researchers need to determine the final high entropy alloy components based on experimental verification. - -## Supported Networks - -| Function | Model | Training | Inferring | Back-end | -|:---------------------| :-------------------- | :--- | :--- |:-----------| -| molecular generation | [high_entropy_alloy_design](https://gitee.com/mindspore/mindscience/tree/master/MindChemistry/applications/high_entropy_alloy_design) | √ | √ | Ascend | diff --git a/docs/mindchemistry/docs/source_en/user/structure_generation.md b/docs/mindchemistry/docs/source_en/user/structure_generation.md new file mode 100644 index 0000000000000000000000000000000000000000..268d294c06fc6e0a80cd6ebe6d448f1051cd5339 --- /dev/null +++ b/docs/mindchemistry/docs/source_en/user/structure_generation.md @@ -0,0 +1,11 @@ +# Structure Generation + +[![View Source On Gitee](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/website-images/master/resource/_static/logo_source_en.svg)](https://gitee.com/mindspore/docs/blob/master/docs/mindchemistry/docs/source_zh_cn/user/structure_generation.md) + +Structure generation, which is a structure generation model based on deep learning to predict the structures of crystalline materials. DiffCSP integrates graph neural networks and equivalent diffusion models to jointly generate crystal lattices and atomic coordinates. It also leverages a periodic E(3)-equivalent denouncing model to better simulate the geometric properties of crystals. Compared with traditional methods based on density functional theory, DiffCSP significantly reduces computational costs and demonstrates excellent performance in crystal structure prediction tasks. + +## Supported Networks + +| Function | Model | Training | Inferring | Back-end | +|:---------------------| :-------------------- | :--- | :--- |:-----------| +| structure generation | [DiffCSP](https://gitee.com/mindspore/mindscience/tree/master/MindChemistry/applications/diffcsp) | √ | √ | Ascend | diff --git a/docs/mindchemistry/docs/source_zh_cn/images/high-alloy_cn.png b/docs/mindchemistry/docs/source_zh_cn/images/high-alloy_cn.png deleted file mode 100644 index e32404974ee5f242b5c64e506aae655c351f2929..0000000000000000000000000000000000000000 Binary files a/docs/mindchemistry/docs/source_zh_cn/images/high-alloy_cn.png and /dev/null differ diff --git a/docs/mindchemistry/docs/source_zh_cn/images/mindchemistry_archi_cn.png b/docs/mindchemistry/docs/source_zh_cn/images/mindchemistry_archi_cn.png index 52b48ff4b6f99a59285fb8addc6ea99a8afd21cc..546a2520923c987d34e0ea2c308b0d8a5001ae2e 100644 Binary files a/docs/mindchemistry/docs/source_zh_cn/images/mindchemistry_archi_cn.png and b/docs/mindchemistry/docs/source_zh_cn/images/mindchemistry_archi_cn.png differ diff --git a/docs/mindchemistry/docs/source_zh_cn/index.rst b/docs/mindchemistry/docs/source_zh_cn/index.rst index 532661d79cd3deab84c55194788b49e198120980..8b69dca3e4885d733c8f7176f92548e6ca5b409b 100644 --- a/docs/mindchemistry/docs/source_zh_cn/index.rst +++ b/docs/mindchemistry/docs/source_zh_cn/index.rst @@ -1,38 +1,28 @@ MindSpore Chemistry文档 -======================= +========================= 介绍 ---- 传统化学研究长期以来面临着众多挑战，实验设计、合成、表征和分析的过程往往耗时、昂贵，并且高度依赖专家经验。AI与化学的协同可以克服传统方法的局限性、开拓全新的研究范式，结合AI模型与化学知识，可以高效处理大量数据、挖掘隐藏的关联信息，构建仿真模型，从而加快化学反应的设计和优化，实现材料的性质预测，并辅助设计新材料。 -**MindSpore -Chemistry**\ （MindChemistry）是基于MindSpore构建的化学领域套件，支持多体系（有机/无机/复合材料化学）、多尺度任务（微观分子生成/预测、宏观反应优化）的AI+化学仿真，致力于高效使能AI与化学的融合研究，践行和牵引AI与化学联合多研究范式跃迁，为化学领域专家的研究提供全新视角与高效的工具。 +**MindSpore Chemistry**\ （MindChemistry）是基于MindSpore构建的化学领域套件，支持多体系（有机/无机/复合材料化学）、多尺度任务（微观分子生成/预测、宏观反应优化）的AI+化学仿真，致力于高效使能AI与化学的融合研究，践行和牵引AI与化学联合多研究范式跃迁，为化学领域专家的研究提供全新视角与高效的工具。 .. figure:: ./images/mindchemistry_archi_cn.png :alt: MindSpore Chemistry Architecture 最新消息 -------- - -- ``2024.07.30`` 2024年7月30日 MindChemistry 0.1.0版本发布。 +- `2025.03.30` MindChemistry 0.2.0版本发布，包括多个应用案例，支持NequIP、Allegro、DeephE3nn、Matformer以及DiffCSP模型。 +- `2024.07.30` MindChemistry 0.1.0版本发布。 特性 ----- +----- 应用案例 ~~~~~~~~ -- **分子生成**\ ： - - - **体系**\ ：无机化学 - - **数据**\ ：高熵合金数据集。高熵合金数据集中包含了已知高熵合金的组分以及热动力学性质等信息，提供金属组分类型及组分比例，以及居里温度、磁致伸缩等热动力学性质信息。 - - **任务**\ ：高熵合金组分设计。我们集成了基于主动学习进行高熵合金设计的方法[1]，设计热膨胀系数极低的高熵合金组分。在主动学习流程中，首先基于AI模型生成候选的高熵合金组分，并基于预测模型和热动力学计算预测热膨胀系数对候选组分进行筛选，最终需要研究者基于实验验证确定最终的高熵合金组分。 - -.. figure:: ./images/high-alloy_cn.png - :alt: MindSpore high alloy Architecture - -- **分子预测**\ ： +- **力场模拟**\ ： - **体系**\ ：有机化学 - **数据**\ ：Revised Molecular Dynamics @@ -42,29 +32,29 @@ Chemistry**\ （MindChemistry）是基于MindSpore构建的化学领域套件， .. figure:: ./images/nequip_cn.png :alt: MindSpore nequip Architecture -- **电子结构预测**\ ： +- **DFT模拟**\ ： - **体系**\ ：材料化学 - **数据**\ ：双层石墨烯数据集。该数据集包含了原子位置、原子数等描述信息以及哈密顿量等性质信息。 - **任务**\ ：密度泛函理论哈密顿量预测。我们集成了DeephE3nn模型[4]，基于E3的等变神经网络，利用原子的结构去预测其的哈密顿量。 -- **晶体材料性质预测**\ ： +- **性质预测**\ ： - **体系**\ ：材料化学 - **数据**\ ：JARVIS-DFT 3D数据集。该数据集包含了晶体材料的原子位置、原子数等描述信息以及能量、力场等性质信息。 - **任务**\ ：晶体材料性质预测。我们集成了Matformer模型[5]，基于图神经网络和Transformer架构的模型，用于预测晶体材料的各种性质。 -功能模块 -~~~~~~~~ - -- **等变计算库** +- **结构生成**\ ： - - **简介**\ ：对称性是科学领域的重要性质。等变神经网络以具有物理意义表征刻画化合物体系输入，并使得输入与输出在空间平移、旋转和反演等变换中具有等变性。使用等变神经网络来对科学场景建模可以提高数据的表征效率和模型的训练效率。 - - **核心模块**\ ：等变计算库中集成了不可约表示、球谐函数以及张量积等基础模块，实现底层逻辑与运算过程，并基于基础模块构建了等变激活层、等变线性层和等变卷积层等神经网络层，可以更方便地调用从而构建等变神经网络。 + - **体系**：材料化学 + - **数据**： + - **Perov-5**：钙钛矿数据集，每个晶胞中固定5个原子，结构接近。 + - **Carbon-24**：碳晶体数据集，包含6到24个碳原子的不同结构。 + - **MP-20**：MP数据集中的实验数据，胞内不超过20个原子。 + - **MPTS-52**：MP-20的进阶版，胞内最多52个原子。 + - **任务**：晶体材料结构预测。集成了 **DiffCSP** 模型[5]，基于图神经网络和扩散模型，预测晶体材料的结构。 -.. figure:: ./images/e3_cn.png - :alt: MindSpore e3 Architecture 安装教程 -------- @@ -136,8 +126,7 @@ master master >=2.3 >=3.8 感谢以下开发者做出的贡献： -yufan, wangzidong, liuhongsheng, gongyue, gengchenhua, linghejing, -yanchaojie, suyun, wujian, caowenbin +wujian, wangyuheng, Lin Peijia, gengchenhua, caowenbin, Siyu Yang 贡献指南 -------- @@ -152,25 +141,15 @@ yanchaojie, suyun, wujian, caowenbin 引用 ---- -[1] Rao Z, Tung P Y, Xie R, et al. Machine learning-enabled high-entropy -alloy discovery[J]. Science, 2022, 378(6615): 78-85. +[1] Batzner S, Musaelian A, Sun L, et al. E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials[J]. Nature communications, 2022, 13(1): 2453. -[2] Batzner S, Musaelian A, Sun L, et al. E(3)-equivariant graph neural -networks for data-efficient and accurate interatomic potentials[J]. -Nature communications, 2022, 13(1): 2453. +[2] Musaelian A, Batzner S, Johansson A, et al. Learning local equivariant representations for large-scale atomistic dynamics[J]. Nature communications, 2023, 14(1): 579. -[3] Musaelian A, Batzner S, Johansson A, et al. Learning local -equivariant representations for large-scale atomistic dynamics[J]. -Nature communications, 2023, 14(1): 579. +[3] Xiaoxun Gong, He Li, Nianlong Zou, et al. General framework for E(3)-equivariant neural network representation of density functional theory Hamiltonian[J]. Nature communications, 2023, 14: 2848. -[4] Xiaoxun Gong, He Li, Nianlong Zou, et al. General framework for -E(3)-equivariant neural network representation of density functional -theory Hamiltonian[J]. -Nature communications, 2023, 14: 2848. +[4] Keqiang Yan, Yi Liu, Yuchao Lin, Shuiwang ji, et al. Periodic Graph Transformers for Crystal Material Property Prediction[J]. arXiv:2209.11807v1 [cs.LG] 23 sep 2022. -[5] Keqiang Yan, Yi Liu, Yuchao Lin, Shuiwang ji, et al. Periodic -Graph Transformers for Crystal Material Property Prediction[J]. -arXiv:2209.11807v1 [cs.LG] 23 sep 2022. +[5] Jiao Rui and Huang Wenbing and Lin Peijia, et al. Crystal structure prediction by joint equivariant diffusion[J]. Advances in Neural Information Processing Systems, 2024, 36. .. toctree:: :maxdepth: 1 @@ -185,7 +164,7 @@ arXiv:2209.11807v1 [cs.LG] 23 sep 2022. :caption: 使用者指南 :hidden: - user/molecular_generation + user/structure_generation user/molecular_prediction .. toctree:: diff --git a/docs/mindchemistry/docs/source_zh_cn/user/molecular_generation.md b/docs/mindchemistry/docs/source_zh_cn/user/structure_generation.md similarity index 33% rename from docs/mindchemistry/docs/source_zh_cn/user/molecular_generation.md rename to docs/mindchemistry/docs/source_zh_cn/user/structure_generation.md index 28a0e275c8a535b25e235ac9413bf1c7ee8ee330..b8c9918776f2661c0c1a0c752d732cd116015d4b 100644 --- a/docs/mindchemistry/docs/source_zh_cn/user/molecular_generation.md +++ b/docs/mindchemistry/docs/source_zh_cn/user/structure_generation.md @@ -1,11 +1,11 @@ -# 分子生成 +# 结构生成 -[![查看源文件](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/website-images/master/resource/_static/logo_source.svg)](https://gitee.com/mindspore/docs/blob/master/docs/mindchemistry/docs/source_zh_cn/user/molecular_generation.md) +[![查看源文件](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/website-images/master/resource/_static/logo_source.svg)](https://gitee.com/mindspore/docs/blob/master/docs/mindchemistry/docs/source_zh_cn/user/structure_generation.md) -分子生成，通过深度学习的生成模型去预测并生成粒子体系中的组成. 我们集成了基于主动学习进行高熵合金设计的方法，设计热膨胀系数极低的高熵合金组分。在主动学习流程中，首先基于AI模型生成候选的高熵合金组分，并基于预测模型和热动力学计算预测热膨胀系数对候选组分进行筛选，最终需要研究者基于实验验证确定最终的高熵合金组分。 +结构生成，通过深度学习的生成模型预测晶体材料的结构。我们集成了基于图神经网络和等变扩散模型的晶体生成模型 DiffCSP，它通过联合生成晶格和原子坐标来预测晶体结构，并利用周期性 E(3) 等变去噪模型来更好地模拟晶体的几何特性。它在计算成本上远低于传统的基于密度泛函理论的方法，且在晶体结构预测任务中表现出色。 ## 已支持网络 | 功能 | 模型 | 训练 | 推理 | 后端 | | :------------- | :-------------------- | :--- | :--- | :-------- | -| 分子生成| [high_entropy_alloy_design](https://gitee.com/mindspore/mindscience/tree/master/MindChemistry/applications/high_entropy_alloy_design) | √ | √ | Ascend | +| 结构生成| [DiffCSP](https://gitee.com/mindspore/mindscience/tree/master/MindChemistry/applications/diffcsp) | √ | √ | Ascend |