Dec 19, 2024 · Pre-trained on a diverse array of chemical and materials systems using a multi-task approach, DPA-2 demonstrates superior generalization capabilities across multiple downstream tasks...

Dec 24, 2023 · Pre-trained on a diverse array of chemical and materials systems using a multi-task approach, DPA-2 demonstrates superior generalization capabilities across multiple downstream tasks compared to the traditional single-task …

Pre-trained on a diverse array of chemical and materials systems using a multi-task approach, DPA-2 demonstrates superior generalization capabilities across multiple downstream tasks compared...

2-Nitrodiphenylamine is an organic chemical with the formula C6H5NHC6H4NO2. It is a nitrated derivative of diphenylamine. It is a red solid, usually found in form of flakes or powder. It is polar but hydrophobic. Diphenylamine is used to extend the shelf-life of explosives containing nitrocellulose or nitroglycerin.

一方面是DPA-2使用了多任务训练策略（multi-task training），可以同时使用DFT设置不同的多个数据集进行预训练，在对下游任务进行微调（finetune）时，模型的主干（将构型空间和化学空间的表示进行编码的部分）将被保留，并在后边连接到一个或多个头部网络。 因此，预训练的数据集和微调的数据集的标注方法不必完全相同。 经过微调得到的模型参数量较多，直接应用于生产场景（如MD模拟）时可能会导致效率低下，为了解决这一问题，我们可以将模型蒸 …

Pre-trained on a diverse array of chemical and materials systems using a multi-task approach, DPA-2 demonstrates superior generalization capabilities across multiple downstream tasks compared to the traditional single-task pre-training and fine-tuning methodologies.

We introduce DPA-2, a sophisticated architecture designed for Large Atomic Models (LAMs), complemented by a comprehensive pipeline encompassing pre-training, fine-tuning, knowledge distillation, and practical deployment.

The DPA-2 model implementation. See DPA-2 paper for more details. Training example: examples/water/dpa2/input_torch_medium.json , see README for inputs in different levels.

Apr 23, 2024 · We show that DPA-2 can accurately represent a diverse range of chemical systems and materials, enabling high-quality simulations and predictions with significantly reduced efforts compared to traditional methods.

First of all, we are excited to announce the beta3 version of DeePMD-kit v3. DeePMD-kit v3 allows you to train and run deep potential models on top of TensorFlow or PyTorch. DeePMD-kit v3 also supports the DPA-2 model, a novel architecture for large atomic models.