3. Distilling a Cyclohexane Potential from a DPA3 Pretrained Model#
This tutorial demonstrates how to distill a compact Deep Potential model for cyclohexane from a DPA3 pretrained Teacher model. It is adapted from the experimental materials of the Artificial Intelligence and Molecular Science course at the University of Science and Technology of China.
The workflow uses the pretrained DPA-3.2-5M model as the Teacher potential. We first run molecular dynamics (MD) for cyclohexane with the Teacher model, then extract configurations from the trajectory together with the Teacher-predicted energy and force labels. Finally, we train a smaller Student DeePMD model that reproduces the Teacher response within the sampled cyclohexane configuration space and use the Student model for a longer MD simulation.
DPA3: a DeePMD-kit atomic environment representation that combines atomic, radial, and angular information to describe local chemical environments.
Knowledge distillation: a model-compression strategy. In this tutorial, the Student model learns Teacher-predicted energy and force labels, not direct DFT labels. The goal is to approximate the Teacher potential for a specific molecular system at a lower inference cost.
Workflow:
Generate an initial cyclohexane structure.
Run Teacher MD with the DPA3 pretrained model.
Extract configurations and Teacher energy / force labels.
Build a DeePMD training dataset.
Train a compact Student potential.
Run longer MD with the Student potential.
Cyclohexane is a convenient documentation example because it is a small organic molecule with only C and H atoms, so the calculation is fast enough for a tutorial. However, a short unbiased MD trajectory started from one initial structure should not be interpreted as a complete conformational search. At room temperature it will usually remain in the chair basin. Observing rare ring-flip events or a statistically meaningful population of boat and twist-boat conformers requires much longer trajectories, higher-temperature sampling, or enhanced-sampling methods. This tutorial therefore uses cyclohexane mainly to demonstrate the distillation workflow and to visualize representative structures within the sampled region.
3.1. Install the required packages#
DeePMD-kit supports multiple backends. Because the pretrained weights used here are based on PyTorch, this tutorial installs the PyTorch backend.
!pip install deepmd-kit[torch] dpdata ase matplotlib openbabel-wheel py3Dmol tqdm
3.2. Import dependencies#
This section imports the Python packages used later in the tutorial. The imports are intentionally kept close to the beginning so that missing packages fail early.
Standard library modules:
json: read and write JSON configuration files.os: query system information and paths.subprocess: run external commands when needed.time: measure elapsed time.pathlib.Path: handle paths in a platform-independent way.
Third-party packages:
numpy: numerical arrays and analysis.dpdata: data conversion and DeePMD dataset writing.
from pathlib import Path
import numpy as np
import dpdata
3.3. Prepare the Teacher model weights#
The distillation workflow requires a pretrained model as the Teacher potential. This tutorial uses the publicly available DPA-3.2-5M model.
Run the following cell to download the weights and copy them into a local weights/ directory.
%%bash
set -euo pipefail
dp pretrained download DPA-3.2-5M
mkdir -p weights
cp "$HOME/.cache/deepmd/pretrained/models/DPA-3.2-5M.pt" "weights/DPA-3.2-5M.pt"
[2026-06-20 21:19:18,866] DEEPMD INFO Selecting fastest source among 3 candidates...
[2026-06-20 21:19:18,866] DEEPMD INFO Downloading 'DPA-3.2-5M' (source 1/3): https://modelscope.cn/models/DeepModelingCommunity/DPA-3.2-5M/resolve/master/DPA-3.2-5M.pt
[2026-06-20 21:19:23,383] DEEPMD INFO Downloaded 'DPA-3.2-5M' to: /root/.cache/deepmd/pretrained/models/DPA-3.2-5M.pt
[2026-06-20 21:19:23,383] DEEPMD INFO Pretrained model path: /root/.cache/deepmd/pretrained/models/DPA-3.2-5M.pt
/root/.cache/deepmd/pretrained/models/DPA-3.2-5M.pt
Inspect the available model branches and the type map.
DPA-3.2-5M is a multitask model. The cells below first list the available branches and type map. The Teacher MD later uses the OMol25 branch explicitly because cyclohexane is an organic molecular system.
!dp --pt show weights/DPA-3.2-5M.pt model-branch type-map
[2026-06-20 21:19:30,456] DEEPMD WARNING To get the best performance, it is recommended to adjust the number of threads by setting the environment variables OMP_NUM_THREADS, DP_INTRA_OP_PARALLELISM_THREADS, and DP_INTER_OP_PARALLELISM_THREADS. See https://deepmd.rtfd.io/parallelism/ for more information.
[2026-06-20 21:19:37,012] DEEPMD WARNING You can use the environment variable DP_INFER_BATCH_SIZE tocontrol the inference batch size (nframes * natoms). The default value is 1024.
[2026-06-20 21:19:40,636] DEEPMD WARNING You can use the environment variable DP_INFER_BATCH_SIZE tocontrol the inference batch size (nframes * natoms). The default value is 1024.
[2026-06-20 21:19:40,636] DEEPMD INFO This is a multitask model
[2026-06-20 21:19:40,637] DEEPMD INFO Available model branches are ['OMat24', 'OMol25', 'OC20M', 'Alloy_APEX', 'Domains_Alloy', 'Alex2D', 'ODAC23', 'Organic_Reactions', 'OC22', 'MPTrj', 'SSE_ABACUS', 'Domains_SSE_PBE', 'Electrolyte', 'Domains_SemiCond', 'Domains_Anode', 'Domains_Cluster', 'Hybrid_Perovskite', 'Domains_FerroEle', 'H2O_H2O_PD', 'Others_In2Se3', 'Metals_AlMgCu', 'Metals_AgAu_PBED3', 'MPGen_OpenCSP', 'RANDOM'], where 'RANDOM' means using a randomly initialized fitting net.
[2026-06-20 21:19:40,641] DEEPMD INFO Detailed information:
+-------------------+---------------------------+--------------------------------+--------------------------------+
| Model Branch | Alias | description | observed_type |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| OMat24 | Default, Materials, | OMat24 is a large-scale open | ['H', 'He', 'Li', 'Be', 'B', |
| | Omat24, materials, omat24 | dataset containing over 110 | 'C', 'N', 'O', 'F', 'Ne', |
| | | million DFT calculations | 'Na', 'Mg', 'Al', 'Si', 'P', |
| | | spanning diverse structures | 'S', 'Cl', 'Ar', 'K', 'Ca', |
| | | and compositions. It is | 'Sc', 'Ti', 'V', 'Cr', 'Mn', |
| | | designed to support AI-driven | 'Fe', 'Co', 'Ni', 'Cu', 'Zn', |
| | | materials discovery by | 'Ga', 'Ge', 'As', 'Se', 'Br', |
| | | providing broad and deep | 'Kr', 'Rb', 'Sr', 'Y', 'Zr', |
| | | coverage of chemical space. | 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', |
| | | | 'Pd', 'Ag', 'Cd', 'In', 'Sn', |
| | | | 'Sb', 'Te', 'I', 'Xe', 'Cs', |
| | | | 'Ba', 'La', 'Ce', 'Pr', 'Nd', |
| | | | 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', |
| | | | 'Dy', 'Ho', 'Er', 'Tm', 'Yb', |
| | | | 'Lu', 'Hf', 'Ta', 'W', 'Re', |
| | | | 'Os', 'Ir', 'Pt', 'Au', 'Hg', |
| | | | 'Tl', 'Pb', 'Bi', 'Ac', 'Th', |
| | | | 'Pa', 'U', 'Np', 'Pu'] |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| OMol25 | Molecules, molecules, | Open Molecules 2025 (OMol25) | ['H', 'He', 'Li', 'Be', 'B', |
| | omol25 | Dataset is a large-scale | 'C', 'N', 'O', 'F', 'Ne', |
| | | resource for training | 'Na', 'Mg', 'Al', 'Si', 'P', |
| | | molecular chemistry machine | 'S', 'Cl', 'Ar', 'K', 'Ca', |
| | | learning models. OMol25 | 'Sc', 'Ti', 'V', 'Cr', 'Mn', |
| | | comprises over 100 million DFT | 'Fe', 'Co', 'Ni', 'Cu', 'Zn', |
| | | single-point calculations | 'Ga', 'Ge', 'As', 'Se', 'Br', |
| | | containing up to 350 atoms at | 'Kr', 'Rb', 'Sr', 'Y', 'Zr', |
| | | a high level of DFT theory. | 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', |
| | | | 'Pd', 'Ag', 'Cd', 'In', 'Sn', |
| | | | 'Sb', 'Te', 'I', 'Xe', 'Cs', |
| | | | 'Ba', 'La', 'Ce', 'Pr', 'Nd', |
| | | | 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', |
| | | | 'Dy', 'Ho', 'Er', 'Tm', 'Yb', |
| | | | 'Lu', 'Hf', 'Ta', 'W', 'Re', |
| | | | 'Os', 'Ir', 'Pt', 'Au', 'Hg', |
| | | | 'Tl', 'Pb', 'Bi'] |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| OC20M | Catalysis, catalysis | The OC20M is a subset of OC20 | ['H', 'B', 'C', 'N', 'O', |
| | | dataset, which contains over | 'Na', 'Al', 'Si', 'P', 'S', |
| | | 1.2 million DFT relaxations | 'Cl', 'K', 'Ca', 'Sc', 'Ti', |
| | | and approximately 265 million | 'V', 'Cr', 'Mn', 'Fe', 'Co', |
| | | single-point evaluations | 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', |
| | | covering diverse catalyst | 'As', 'Se', 'Rb', 'Sr', 'Y', |
| | | surfaces and adsorbates | 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', |
| | | involving C, N, and O species. | 'Rh', 'Pd', 'Ag', 'Cd', 'In', |
| | | | 'Sn', 'Sb', 'Te', 'Cs', 'Hf', |
| | | | 'Ta', 'W', 'Re', 'Os', 'Ir', |
| | | | 'Pt', 'Au', 'Hg', 'Tl', 'Pb', |
| | | | 'Bi'] |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| Alloy_APEX | Alloys, Alloy_tongqi, | The dataset covers elemental | ['Li', 'Be', 'Na', 'Mg', 'Al', |
| | Li2025APEX, alloys | and multi-component alloy | 'Si', 'K', 'Ca', 'Sc', 'Ti', |
| | | systems composed of 53 | 'V', 'Cr', 'Mn', 'Fe', 'Co', |
| | | metallic elements, including | 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', |
| | | structures with defects such | 'Sr', 'Y', 'Zr', 'Nb', 'Mo', |
| | | as vacancies, interstitials, | 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', |
| | | and surfaces. These | 'In', 'Sn', 'La', 'Ce', 'Pr', |
| | | configurations are generated | 'Nd', 'Sm', 'Gd', 'Tb', 'Dy', |
| | | for single elements, | 'Ho', 'Er', 'Tm', 'Lu', 'Hf', |
| | | compounds, solid solutions, | 'Ta', 'W', 'Re', 'Os', 'Ir', |
| | | and their defects over 50–3000 | 'Pt', 'Au', 'Pb'] |
| | | K and −0.5–5 GPa. | |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| Domains_Alloy | Dai2023Alloy | The dataset contains | ['Li', 'Be', 'Na', 'Mg', 'Al', |
| | | structure-energy-force-virial | 'Si', 'K', 'Ca', 'Sc', 'Ti', |
| | | data for 53 typical metallic | 'V', 'Cr', 'Mn', 'Fe', 'Co', |
| | | elements in alloy systems, | 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', |
| | | including ~9000 intermetallic | 'Sr', 'Y', 'Zr', 'Nb', 'Mo', |
| | | compounds and FCC, BCC, HCP | 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', |
| | | structures. It consists of two | 'In', 'Sn', 'La', 'Ce', 'Pr', |
| | | parts: DFT-generated relaxed | 'Nd', 'Sm', 'Gd', 'Tb', 'Dy', |
| | | and deformed structures, and | 'Ho', 'Er', 'Tm', 'Lu', 'Hf', |
| | | randomly distorted structures | 'Ta', 'W', 'Re', 'Os', 'Ir', |
| | | produced covering pure metals, | 'Pt', 'Au', 'Pb'] |
| | | solid solutions, and | |
| | | intermetallics with vacancies. | |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| Alex2D | 2DMaterials, 2dmaterials | This dataset contains | ['H', 'Li', 'Be', 'B', 'C', |
| | | approximately 6,500 novel two- | 'N', 'O', 'F', 'Na', 'Mg', |
| | | dimensional materials | 'Al', 'Si', 'P', 'S', 'Cl', |
| | | generated through a symmetry- | 'K', 'Ca', 'Sc', 'Ti', 'V', |
| | | based combinatorial method | 'Cr', 'Mn', 'Fe', 'Co', 'Ni', |
| | | that systematically fills | 'Cu', 'Zn', 'Ga', 'Ge', 'As', |
| | | Wyckoff positions under | 'Se', 'Br', 'Rb', 'Sr', 'Y', |
| | | constraints of charge | 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', |
| | | neutrality and | 'Rh', 'Pd', 'Ag', 'Cd', 'In', |
| | | electronegativity balance. The | 'Sn', 'Sb', 'Te', 'I', 'Cs', |
| | | resulting structures span over | 'Ba', 'La', 'Ce', 'Pr', 'Nd', |
| | | 30 stoichiometries, exhibit | 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', |
| | | diverse tiling patterns and | 'Dy', 'Ho', 'Er', 'Tm', 'Yb', |
| | | polymorphisms, and all lie | 'Lu', 'Hf', 'Ta', 'W', 'Re', |
| | | within 250 meV/atom of the | 'Os', 'Ir', 'Pt', 'Au', 'Hg', |
| | | thermodynamic convex hull. | 'Tl', 'Pb', 'Bi', 'Ac', 'Th', |
| | | | 'Pa', 'U', 'Np', 'Pu'] |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| ODAC23 | | The dataset contains over 38 | ['H', 'Li', 'Be', 'B', 'C', |
| | | million quantum chemistry | 'N', 'O', 'F', 'Na', 'Mg', |
| | | calculations on thousands of | 'Al', 'Si', 'P', 'S', 'Cl', |
| | | metal-organic frameworks | 'Ca', 'Sc', 'Ti', 'V', 'Cr', |
| | | (MOFs) interacting with carbon | 'Mn', 'Fe', 'Co', 'Ni', 'Cu', |
| | | dioxide and water. It provides | 'Zn', 'Ga', 'Ge', 'As', 'Se', |
| | | comprehensive data to support | 'Br', 'Sr', 'Y', 'Zr', 'Nb', |
| | | machine learning-driven | 'Mo', 'Ru', 'Rh', 'Pd', 'Ag', |
| | | development of MOFs for direct | 'Cd', 'Sn', 'Sb', 'Te', 'I', |
| | | air capture (DAC) | 'Cs', 'Ba', 'La', 'Ce', 'Pr', |
| | | applications. | 'Nd', 'Sm', 'Eu', 'Gd', 'Tb', |
| | | | 'Dy', 'Ho', 'Er', 'Tm', 'Lu', |
| | | | 'Hf', 'W', 'Re', 'Pt', 'Au', |
| | | | 'Hg', 'Bi', 'Th', 'U', 'Np'] |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| Organic_Reactions | Li2025General | The dataset consists of over | ['H', 'C', 'N', 'O'] |
| | | 17 million semi-empirical | |
| | | energy-labeled non-equilibrium | |
| | | structures along reaction | |
| | | pathways involving C, H, O, | |
| | | and N, generated using NEB and | |
| | | structural alignment. It is | |
| | | complemented by a fine-tuning | |
| | | dataset of over 200,000 DFT- | |
| | | labeled structures selected | |
| | | via active learning to support | |
| | | the development of reactive | |
| | | machine learning potentials. | |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| OC22 | | The OC22 dataset contains DFT | ['H', 'Li', 'Be', 'C', 'N', |
| | | relaxation trajectories and | 'O', 'Na', 'Mg', 'Al', 'Si', |
| | | single-point calculations | 'K', 'Ca', 'Sc', 'Ti', 'V', |
| | | covering a diverse set of | 'Cr', 'Mn', 'Fe', 'Co', 'Ni', |
| | | oxide materials, adsorbates, | 'Cu', 'Zn', 'Ga', 'Ge', 'As', |
| | | and coverages relevant to | 'Se', 'Rb', 'Sr', 'Y', 'Zr', |
| | | Oxygen Evolution Reaction | 'Nb', 'Mo', 'Ru', 'Rh', 'Pd', |
| | | catalysts. It provides a | 'Ag', 'Cd', 'In', 'Sn', 'Sb', |
| | | large-scale, open benchmark | 'Te', 'Cs', 'Ba', 'Ce', 'Lu', |
| | | for training machine learning | 'Hf', 'Ta', 'W', 'Re', 'Os', |
| | | models on total energy and | 'Ir', 'Pt', 'Au', 'Hg', 'Tl', |
| | | forces predictions for oxide | 'Pb', 'Bi'] |
| | | electrocatalysts. | |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| MPTrj | MP_traj_v024_alldata_mixu | The MPtrj dataset contains DFT | ['H', 'He', 'Li', 'Be', 'B', |
| | | trajectory data for 145,923 | 'C', 'N', 'O', 'F', 'Ne', |
| | | compounds from the Materials | 'Na', 'Mg', 'Al', 'Si', 'P', |
| | | Project, curated by filtering | 'S', 'Cl', 'Ar', 'K', 'Ca', |
| | | GGA and GGA+U calculations for | 'Sc', 'Ti', 'V', 'Cr', 'Mn', |
| | | consistency, convergence, and | 'Fe', 'Co', 'Ni', 'Cu', 'Zn', |
| | | energy quality. It includes | 'Ga', 'Ge', 'As', 'Se', 'Br', |
| | | non-deprecated, non-duplicate | 'Kr', 'Rb', 'Sr', 'Y', 'Zr', |
| | | structures with verified | 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', |
| | | settings, enabling reliable | 'Pd', 'Ag', 'Cd', 'In', 'Sn', |
| | | machine learning on energy and | 'Sb', 'Te', 'I', 'Xe', 'Cs', |
| | | force predictions across a | 'Ba', 'La', 'Ce', 'Pr', 'Nd', |
| | | broad materials space. | 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', |
| | | | 'Dy', 'Ho', 'Er', 'Tm', 'Yb', |
| | | | 'Lu', 'Hf', 'Ta', 'W', 'Re', |
| | | | 'Os', 'Ir', 'Pt', 'Au', 'Hg', |
| | | | 'Tl', 'Pb', 'Bi', 'Ac', 'Th', |
| | | | 'Pa', 'U', 'Np', 'Pu'] |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| SSE_ABACUS | Shi2024SSE | This dataset can be used to | ['Li', 'B', 'O', 'Al', 'Si', |
| | | study and predict the | 'P', 'S', 'Cl', 'Sc', 'Ga', |
| | | properties and behavior of | 'Ge', 'As', 'Se', 'Br', 'Y', |
| | | solid-state electrolytes under | 'Zr', 'In', 'Sn', 'Sb', 'I', |
| | | various conditions, such as | 'Dy', 'Ho', 'Er', 'Tm', 'Yb', |
| | | different temperatures and | 'Lu', 'Ta'] |
| | | pressures. This can help | |
| | | researchers and engineers | |
| | | design better materials for | |
| | | use in energy storage devices, | |
| | | such as batteries and | |
| | | supercapacitors. | |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| Domains_SSE_PBE | Huang2021Deep-PBE | The dataset consists of | ['Li', 'Si', 'P', 'S', 'Ge', |
| | | interatomic potentials and | 'Sn'] |
| | | simulation data for | |
| | | Li10GeP2S12-type solid-state | |
| | | electrolytes, including | |
| | | Li10GeP2S12, Li10SiP2S12, and | |
| | | Li10SnP2S12. It covers | |
| | | diffusion processes across a | |
| | | wide temperature range and | |
| | | large system sizes, | |
| | | incorporating effects of | |
| | | thermal expansion, | |
| | | configurational disorder, and | |
| | | density functional variations. | |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| Electrolyte | Shi2024Electrolyte | The dataset includes samples | ['H', 'Li', 'C', 'O', 'F', |
| | | of systems containing Li, P, | 'P'] |
| | | F, C, H, and O, specifically | |
| | | focused on lithium | |
| | | hexafluorophosphate and | |
| | | various carbonate solvents. It | |
| | | covers a wide range of | |
| | | temperatures from 0 to 450 K | |
| | | and pressures from 0 to 1 GPa, | |
| | | with LiPF6 concentrations | |
| | | between 0.8 and 1.2 mol/L. | |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| Domains_SemiCond | Liu2024Machine | For 19 semiconductors ranging | ['B', 'C', 'N', 'Al', 'Si', |
| | | from group IIB to VIA, | 'P', 'S', 'Zn', 'Ga', 'Ge', |
| | | including Si, Ge, SiC, BAs, | 'As', 'Se', 'Cd', 'In', 'Sb', |
| | | BN, AlN, AlP, AlAs, InP, InAs, | 'Te'] |
| | | InSb, GaN, GaP, GaAs, CdTe, | |
| | | InTe, CdSe, ZnS, and CdS. | |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| Domains_Anode | Zhang2023Cathode | The dataset contains O3-type | ['Fe', 'Co', 'Li', 'O', 'Cr', |
| | | layered oxide cathodes | 'Ni', 'Na', 'Mn'] |
| | | (LixTMO2 and NaxTMO2, TM = Ni, | |
| | | Mn, Fe, Co, Cr) generated from | |
| | | ~300 bulk systems. It spans | |
| | | compositions with x = 0, 0.5, | |
| | | 1, includes Jahn-Teller | |
| | | distorted structures, and | |
| | | covers temperatures from 50 K | |
| | | to 1250 K and pressures from 0 | |
| | | to 3000 bar. | |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| Domains_Cluster | Gong2023Cluster | The dataset contains | ['Al', 'Si', 'Ni', 'Cu', 'Ru', |
| | | structure-to-energy-force | 'Pd', 'Ag', 'Pt', 'Au'] |
| | | labels for 31 mono- and multi- | |
| | | metallic clusters generated. | |
| | | It covers diverse elemental | |
| | | combinations relevant to | |
| | | catalysis, namely, Au, Ag, Cu, | |
| | | Pt, Pd, Ni, Si, Al, Ru, AuAg, | |
| | | AuCu, AgCu, AuPt, AgPt, CuPt, | |
| | | AuPd, AgPd, CuPd, AuNi, AgNi, | |
| | | CuNi, PtPd, PtNi, NiPd, AgCuPt | |
| | | AuAgCu, AuAgPd, AuAgPt, | |
| | | AuCuPd, AuCuPt, PtPdNi. | |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| Hybrid_Perovskite | Tuo2023Hybrid | Dataset for lead-based | ['H', 'C', 'N', 'I', 'Pb'] |
| | | organic-inorganic hybrid | |
| | | perovskite MAPbI3 and FAPbI3. | |
| | | The cover temperature range is | |
| | | 50~800K. The covered pressure | |
| | | range is below 1GPa. | |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| Domains_FerroEle | UniPero | The dataset includes | ['O', 'Na', 'Mg', 'K', 'Ca', |
| | | perovskite oxides with | 'Ti', 'Zn', 'Sr', 'Zr', 'Nb', |
| | | increasing chemical | 'In', 'Ba', 'Hf', 'Pb', 'Bi'] |
| | | complexity, ranging from | |
| | | simple three-element systems | |
| | | like PbTiO3, BaTiO3, and | |
| | | SrTiO3 to complex solid | |
| | | solutions such as | |
| | | Pb(Mg1/3Nb2/3)O3 and PIN-PMN- | |
| | | PT. It covers around 200 | |
| | | compositions involving 14 | |
| | | different metal elements. | |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| H2O_H2O_PD | water, Zhang2021Phase | Water phase diagram, covering | ['H', 'O'] |
| | | from low temperature and | |
| | | pressure to about 2400 K and | |
| | | 50 GPa, excluding the vapor | |
| | | stability region. | |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| Others_In2Se3 | Wu2021Accurate | The dataset contains diverse | ['Se', 'In'] |
| | | monolayer α-In2Se3 | |
| | | configurations, which supports | |
| | | the development of a high- | |
| | | accuracy deep neural network | |
| | | potential capable of | |
| | | reproducing thermodynamic | |
| | | properties, polarization | |
| | | switching pathways, domain- | |
| | | wall kinetics, and a | |
| | | temperature-driven phase | |
| | | transition. | |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| Metals_AlMgCu | Jiang2021Accurate | For Al-Mg-Cu alloy. | ['Al', 'Mg', 'Cu'] |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| Metals_AgAu_PBED3 | Wang2021Generalizable | Ag-Au nanoalloys. | ['Ag', 'Au'] |
+-------------------+---------------------------+--------------------------------+--------------------------------+
| MPGen_OpenCSP | OpenCSP, opencsp | The OpenCSP dataset was | ['H', 'He', 'Li', 'Be', 'B', |
| | | constructed by relaxing | 'C', 'N', 'O', 'F', 'Ne', |
| | | CALYPSO-proposed structures to | 'Na', 'Mg', 'Al', 'Si', 'P', |
| | | pressure-constrained local | 'S', 'Cl', 'Ar', 'K', 'Ca', |
| | | minima on the potential energy | 'Sc', 'Ti', 'V', 'Cr', 'Mn', |
| | | surface, ensuring direct | 'Fe', 'Co', 'Ni', 'Cu', 'Zn', |
| | | applicability to CSP tasks. | 'Ga', 'Ge', 'As', 'Se', 'Br', |
| | | All data were obtained through | 'Kr', 'Rb', 'Sr', 'Y', 'Zr', |
| | | single-point ABACUS | 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', |
| | | calculations, with structures | 'Pd', 'Ag', 'Cd', 'In', 'Sn', |
| | | generated using the DP-GEN | 'Sb', 'Te', 'I', 'Xe', 'Cs', |
| | | concurrent learning framework, | 'Ba', 'La', 'Ce', 'Pr', 'Nd', |
| | | which samples relaxation | 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', |
| | | trajectories under randomly | 'Dy', 'Ho', 'Er', 'Tm', 'Yb', |
| | | sampled pressure conditions. | 'Lu', 'Hf', 'Ta', 'W', 'Re', |
| | | | 'Os', 'Ir', 'Pt', 'Au', 'Hg', |
| | | | 'Tl', 'Pb', 'Bi'] |
+-------------------+---------------------------+--------------------------------+--------------------------------+
[2026-06-20 21:19:40,644] DEEPMD INFO The type_map of branch OMat24 is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,644] DEEPMD INFO The type_map of branch OMol25 is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,644] DEEPMD INFO The type_map of branch OC20M is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,644] DEEPMD INFO The type_map of branch Alloy_APEX is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,644] DEEPMD INFO The type_map of branch Domains_Alloy is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,644] DEEPMD INFO The type_map of branch Alex2D is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,644] DEEPMD INFO The type_map of branch ODAC23 is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,644] DEEPMD INFO The type_map of branch Organic_Reactions is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,646] DEEPMD INFO The type_map of branch OC22 is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,646] DEEPMD INFO The type_map of branch MPTrj is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,646] DEEPMD INFO The type_map of branch SSE_ABACUS is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,646] DEEPMD INFO The type_map of branch Domains_SSE_PBE is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,646] DEEPMD INFO The type_map of branch Electrolyte is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,646] DEEPMD INFO The type_map of branch Domains_SemiCond is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,646] DEEPMD INFO The type_map of branch Domains_Anode is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,646] DEEPMD INFO The type_map of branch Domains_Cluster is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,646] DEEPMD INFO The type_map of branch Hybrid_Perovskite is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,646] DEEPMD INFO The type_map of branch Domains_FerroEle is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,646] DEEPMD INFO The type_map of branch H2O_H2O_PD is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,646] DEEPMD INFO The type_map of branch Others_In2Se3 is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,646] DEEPMD INFO The type_map of branch Metals_AlMgCu is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,646] DEEPMD INFO The type_map of branch Metals_AgAu_PBED3 is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
[2026-06-20 21:19:40,646] DEEPMD INFO The type_map of branch MPGen_OpenCSP is ['H', 'He', 'Li', 'Be', 'B', 'C', 'N', 'O', 'F', 'Ne', 'Na', 'Mg', 'Al', 'Si', 'P', 'S', 'Cl', 'Ar', 'K', 'Ca', 'Sc', 'Ti', 'V', 'Cr', 'Mn', 'Fe', 'Co', 'Ni', 'Cu', 'Zn', 'Ga', 'Ge', 'As', 'Se', 'Br', 'Kr', 'Rb', 'Sr', 'Y', 'Zr', 'Nb', 'Mo', 'Tc', 'Ru', 'Rh', 'Pd', 'Ag', 'Cd', 'In', 'Sn', 'Sb', 'Te', 'I', 'Xe', 'Cs', 'Ba', 'La', 'Ce', 'Pr', 'Nd', 'Pm', 'Sm', 'Eu', 'Gd', 'Tb', 'Dy', 'Ho', 'Er', 'Tm', 'Yb', 'Lu', 'Hf', 'Ta', 'W', 'Re', 'Os', 'Ir', 'Pt', 'Au', 'Hg', 'Tl', 'Pb', 'Bi', 'Po', 'At', 'Rn', 'Fr', 'Ra', 'Ac', 'Th', 'Pa', 'U', 'Np', 'Pu', 'Am', 'Cm', 'Bk', 'Cf', 'Es', 'Fm', 'Md', 'No', 'Lr', 'Rf', 'Db', 'Sg', 'Bh', 'Hs', 'Mt', 'Ds', 'Rg', 'Cn', 'Nh', 'Fl', 'Mc', 'Lv', 'Ts', 'Og']
3.4. Prepare the distillation dataset#
This section uses Open Babel to generate an initial cyclohexane structure from a SMILES string. It then runs MD with the DPA3 Teacher model through the ASE calculator interface, samples configurations from the trajectory, and writes a DeePMD-kit training dataset.
The workflow is:
Convert the cyclohexane SMILES string into an initial 3D XYZ structure.
Run a Teacher MD trajectory with the DPA3 pretrained model.
Sample configurations from the trajectory and store Teacher-predicted energy and force labels.
Write the sampled configurations as a DeePMD training dataset.
Produce
coord.npy,box.npy,energy.npy,force.npy,type.raw, andtype_map.raw.
The Student model will learn the Teacher response in the configuration region visited by this MD trajectory.
3.4.1. Generate an initial 3D structure from a SMILES string#
Open Babel converts the cyclohexane SMILES string into an initial 3D XYZ structure. This structure is used as the starting point for the Teacher MD simulation.
initial_smiles = "C1CCCCC1"
initial_xyz = Path("data/cyclohexane/raw/cyclohexane_initial.xyz")
initial_xyz.parent.mkdir(parents=True, exist_ok=True)
!obabel -:"{initial_smiles}" -O {initial_xyz} -h --gen3d
print("initial xyz:", initial_xyz)
1 molecule converted
initial xyz: data/cyclohexane/raw/cyclohexane_initial.xyz
!cat data/cyclohexane/raw/cyclohexane_initial.xyz
18
C -1.42059 0.32798 -0.22055
C -0.99079 -1.06496 0.23414
C 0.42516 -1.39045 -0.23359
C 1.41996 -0.32808 0.22623
C 0.99446 1.06588 -0.22741
C -0.42546 1.39336 0.23081
H -2.41457 0.55707 0.18054
H -1.50452 0.34555 -1.31413
H -1.69074 -1.81365 -0.15344
H -1.03516 -1.12043 1.32884
H 0.44219 -1.45748 -1.32822
H 0.72602 -2.37082 0.15224
H 2.41657 -0.55788 -0.16811
H 1.49594 -0.34832 1.32046
H 1.04706 1.12499 -1.32177
H 1.69167 1.81267 0.16780
H -0.72489 2.37058 -0.16285
H -0.44487 1.46853 1.32500
3.4.2. Run Teacher MD and sample configurations#
DeePMD-kit can be used directly as an ASE calculator. After assigning atoms.calc = DP(...), ASE calls the DeePMD model through standard methods such as get_potential_energy() and get_forces().
DPA-3.2-5M is a multitask model. The Teacher calculation uses head="OMol25", so the labels come from the Open Molecules 2025 branch.
The default trajectory below uses 500 MD steps with a 0.5 fs time step. This keeps the documentation example quick to run while still producing a small local configuration distribution for the Student model. It is not intended to exhaustively sample cyclohexane ring inversion.
Reference: DeePMD-kit ASE calculator interface
teacher_md_model = Path("weights/DPA-3.2-5M.pt")
teacher_md_head = "OMol25"
from ase import units
from ase.io import read
from ase.md.velocitydistribution import MaxwellBoltzmannDistribution
from ase.md.verlet import VelocityVerlet
from deepmd.calculator import DP
RANDOM_SEED = 42
MD_TEMPERATURE_K = 300
MD_TIMESTEP_FS = 0.5
MD_STEPS = 500
atoms_obj = read(str(initial_xyz))
atoms_obj.calc = DP(model=str(teacher_md_model), head=teacher_md_head)
rng = np.random.default_rng(RANDOM_SEED)
MaxwellBoltzmannDistribution(atoms_obj, temperature_K=MD_TEMPERATURE_K, rng=rng)
dynamics = VelocityVerlet(
atoms_obj,
timestep=MD_TIMESTEP_FS * units.fs,
trajectory="md.traj",
logfile="-",
)
print(f"Running Teacher MD sampling for {MD_STEPS} steps...")
print("teacher model:", teacher_md_model)
print("teacher head:", teacher_md_head)
dynamics.run(MD_STEPS)
You can use the environment variable DP_INFER_BATCH_SIZE tocontrol the inference batch size (nframes * natoms). The default value is 1024.
Time[ps] Etot[eV] Epot[eV] Ekin[eV] T[K]
Running Teacher MD sampling for 500 steps...
teacher model: weights/DPA-3.2-5M.pt
teacher head: OMol25
0.0000 -6416.6862 -6417.1005 0.4142 178.0
0.0005 -6416.6862 -6417.0937 0.4076 175.2
0.0010 -6416.6860 -6417.0756 0.3896 167.5
0.0015 -6416.6857 -6417.0492 0.3635 156.2
0.0020 -6416.6855 -6417.0187 0.3332 143.2
0.0025 -6416.6853 -6416.9882 0.3030 130.2
0.0030 -6416.6852 -6416.9611 0.2760 118.6
0.0035 -6416.6852 -6416.9388 0.2536 109.0
0.0040 -6416.6853 -6416.9207 0.2354 101.2
0.0045 -6416.6855 -6416.9051 0.2196 94.4
0.0050 -6416.6856 -6416.8894 0.2039 87.6
0.0055 -6416.6856 -6416.8721 0.1865 80.1
0.0060 -6416.6855 -6416.8526 0.1671 71.8
0.0065 -6416.6853 -6416.8326 0.1472 63.3
0.0070 -6416.6851 -6416.8148 0.1297 55.8
0.0075 -6416.6849 -6416.8030 0.1182 50.8
0.0080 -6416.6848 -6416.8004 0.1156 49.7
0.0085 -6416.6849 -6416.8083 0.1234 53.0
0.0090 -6416.6850 -6416.8258 0.1408 60.5
0.0095 -6416.6853 -6416.8500 0.1647 70.8
0.0100 -6416.6856 -6416.8763 0.1907 82.0
0.0105 -6416.6859 -6416.9000 0.2142 92.0
0.0110 -6416.6860 -6416.9175 0.2316 99.5
0.0115 -6416.6859 -6416.9272 0.2412 103.7
0.0120 -6416.6858 -6416.9294 0.2436 104.7
0.0125 -6416.6856 -6416.9265 0.2409 103.5
0.0130 -6416.6854 -6416.9217 0.2363 101.6
0.0135 -6416.6852 -6416.9183 0.2330 100.2
0.0140 -6416.6852 -6416.9181 0.2329 100.1
0.0145 -6416.6853 -6416.9215 0.2362 101.5
0.0150 -6416.6855 -6416.9272 0.2417 103.9
0.0155 -6416.6856 -6416.9325 0.2468 106.1
0.0160 -6416.6858 -6416.9347 0.2490 107.0
0.0165 -6416.6858 -6416.9321 0.2463 105.9
0.0170 -6416.6857 -6416.9243 0.2386 102.6
0.0175 -6416.6855 -6416.9128 0.2273 97.7
0.0180 -6416.6853 -6416.9006 0.2153 92.5
0.0185 -6416.6851 -6416.8912 0.2061 88.6
0.0190 -6416.6850 -6416.8877 0.2027 87.1
0.0195 -6416.6850 -6416.8917 0.2067 88.8
0.0200 -6416.6852 -6416.9026 0.2174 93.5
0.0205 -6416.6854 -6416.9178 0.2324 99.9
0.0210 -6416.6857 -6416.9334 0.2478 106.5
0.0215 -6416.6858 -6416.9458 0.2600 111.7
0.0220 -6416.6859 -6416.9523 0.2664 114.5
0.0225 -6416.6858 -6416.9520 0.2661 114.4
0.0230 -6416.6856 -6416.9461 0.2605 111.9
0.0235 -6416.6854 -6416.9373 0.2519 108.3
0.0240 -6416.6853 -6416.9288 0.2435 104.7
0.0245 -6416.6852 -6416.9231 0.2379 102.2
0.0250 -6416.6852 -6416.9214 0.2361 101.5
0.0255 -6416.6854 -6416.9231 0.2378 102.2
0.0260 -6416.6855 -6416.9263 0.2408 103.5
0.0265 -6416.6857 -6416.9282 0.2425 104.2
0.0270 -6416.6858 -6416.9260 0.2401 103.2
0.0275 -6416.6858 -6416.9181 0.2323 99.9
0.0280 -6416.6857 -6416.9048 0.2191 94.2
0.0285 -6416.6855 -6416.8879 0.2024 87.0
0.0290 -6416.6852 -6416.8707 0.1854 79.7
0.0295 -6416.6850 -6416.8570 0.1719 73.9
0.0300 -6416.6849 -6416.8498 0.1649 70.9
0.0305 -6416.6850 -6416.8508 0.1658 71.3
0.0310 -6416.6852 -6416.8592 0.1740 74.8
0.0315 -6416.6854 -6416.8725 0.1871 80.4
0.0320 -6416.6856 -6416.8870 0.2013 86.5
0.0325 -6416.6857 -6416.8993 0.2135 91.8
0.0330 -6416.6858 -6416.9072 0.2214 95.2
0.0335 -6416.6857 -6416.9105 0.2248 96.6
0.0340 -6416.6855 -6416.9107 0.2252 96.8
0.0345 -6416.6853 -6416.9105 0.2252 96.8
0.0350 -6416.6852 -6416.9130 0.2278 97.9
0.0355 -6416.6851 -6416.9200 0.2349 100.9
0.0360 -6416.6852 -6416.9321 0.2469 106.1
0.0365 -6416.6854 -6416.9478 0.2624 112.8
0.0370 -6416.6856 -6416.9644 0.2788 119.8
0.0375 -6416.6858 -6416.9786 0.2928 125.8
0.0380 -6416.6859 -6416.9874 0.3015 129.6
0.0385 -6416.6859 -6416.9891 0.3032 130.3
0.0390 -6416.6858 -6416.9839 0.2981 128.1
0.0395 -6416.6856 -6416.9735 0.2879 123.7
0.0400 -6416.6854 -6416.9607 0.2753 118.3
0.0405 -6416.6852 -6416.9487 0.2635 113.2
0.0410 -6416.6852 -6416.9402 0.2550 109.6
0.0415 -6416.6852 -6416.9361 0.2509 107.8
0.0420 -6416.6853 -6416.9359 0.2506 107.7
0.0425 -6416.6855 -6416.9374 0.2518 108.2
0.0430 -6416.6857 -6416.9377 0.2521 108.3
0.0435 -6416.6857 -6416.9347 0.2490 107.0
0.0440 -6416.6857 -6416.9272 0.2415 103.8
0.0445 -6416.6856 -6416.9158 0.2302 98.9
0.0450 -6416.6854 -6416.9026 0.2172 93.4
0.0455 -6416.6852 -6416.8907 0.2055 88.3
0.0460 -6416.6851 -6416.8827 0.1976 84.9
0.0465 -6416.6851 -6416.8804 0.1953 83.9
0.0470 -6416.6852 -6416.8838 0.1986 85.4
0.0475 -6416.6854 -6416.8913 0.2060 88.5
0.0480 -6416.6855 -6416.9005 0.2149 92.4
0.0485 -6416.6857 -6416.9083 0.2226 95.7
0.0490 -6416.6858 -6416.9125 0.2268 97.5
0.0495 -6416.6857 -6416.9120 0.2263 97.3
0.0500 -6416.6856 -6416.9071 0.2215 95.2
0.0505 -6416.6855 -6416.8994 0.2139 91.9
0.0510 -6416.6853 -6416.8913 0.2060 88.5
0.0515 -6416.6852 -6416.8853 0.2001 86.0
0.0520 -6416.6852 -6416.8831 0.1979 85.1
0.0525 -6416.6853 -6416.8850 0.1997 85.8
0.0530 -6416.6854 -6416.8899 0.2045 87.9
0.0535 -6416.6856 -6416.8956 0.2101 90.3
0.0540 -6416.6857 -6416.8997 0.2140 92.0
0.0545 -6416.6857 -6416.9002 0.2145 92.2
0.0550 -6416.6857 -6416.8966 0.2109 90.7
0.0555 -6416.6855 -6416.8898 0.2043 87.8
0.0560 -6416.6853 -6416.8821 0.1968 84.6
0.0565 -6416.6852 -6416.8764 0.1912 82.2
0.0570 -6416.6851 -6416.8752 0.1901 81.7
0.0575 -6416.6851 -6416.8797 0.1946 83.6
0.0580 -6416.6852 -6416.8897 0.2045 87.9
0.0585 -6416.6854 -6416.9037 0.2183 93.8
0.0590 -6416.6855 -6416.9191 0.2336 100.4
0.0595 -6416.6856 -6416.9334 0.2478 106.5
0.0600 -6416.6857 -6416.9448 0.2591 111.4
0.0605 -6416.6857 -6416.9526 0.2669 114.7
0.0610 -6416.6856 -6416.9571 0.2715 116.7
0.0615 -6416.6854 -6416.9598 0.2744 117.9
0.0620 -6416.6853 -6416.9626 0.2773 119.2
0.0625 -6416.6853 -6416.9670 0.2817 121.1
0.0630 -6416.6854 -6416.9735 0.2881 123.8
0.0635 -6416.6855 -6416.9813 0.2958 127.1
0.0640 -6416.6856 -6416.9885 0.3029 130.2
0.0645 -6416.6858 -6416.9926 0.3068 131.9
0.0650 -6416.6859 -6416.9914 0.3055 131.3
0.0655 -6416.6858 -6416.9836 0.2978 128.0
0.0660 -6416.6857 -6416.9697 0.2840 122.0
0.0665 -6416.6855 -6416.9515 0.2659 114.3
0.0670 -6416.6853 -6416.9319 0.2465 106.0
0.0675 -6416.6852 -6416.9140 0.2288 98.3
0.0680 -6416.6851 -6416.9002 0.2150 92.4
0.0685 -6416.6852 -6416.8913 0.2062 88.6
0.0690 -6416.6853 -6416.8870 0.2017 86.7
0.0695 -6416.6854 -6416.8857 0.2003 86.1
0.0700 -6416.6855 -6416.8854 0.1999 85.9
0.0705 -6416.6855 -6416.8845 0.1990 85.5
0.0710 -6416.6855 -6416.8821 0.1966 84.5
0.0715 -6416.6854 -6416.8786 0.1932 83.0
0.0720 -6416.6853 -6416.8752 0.1899 81.6
0.0725 -6416.6852 -6416.8740 0.1888 81.2
0.0730 -6416.6851 -6416.8769 0.1918 82.4
0.0735 -6416.6851 -6416.8851 0.1999 85.9
0.0740 -6416.6852 -6416.8984 0.2132 91.6
0.0745 -6416.6854 -6416.9154 0.2300 98.8
0.0750 -6416.6856 -6416.9332 0.2476 106.4
0.0755 -6416.6858 -6416.9489 0.2631 113.1
0.0760 -6416.6859 -6416.9595 0.2737 117.6
0.0765 -6416.6859 -6416.9638 0.2779 119.4
0.0770 -6416.6857 -6416.9617 0.2759 118.6
0.0775 -6416.6856 -6416.9550 0.2694 115.8
0.0780 -6416.6854 -6416.9463 0.2609 112.1
0.0785 -6416.6853 -6416.9384 0.2531 108.8
0.0790 -6416.6853 -6416.9329 0.2477 106.4
0.0795 -6416.6853 -6416.9304 0.2451 105.3
0.0800 -6416.6854 -6416.9301 0.2446 105.1
0.0805 -6416.6856 -6416.9302 0.2446 105.1
0.0810 -6416.6856 -6416.9289 0.2432 104.5
0.0815 -6416.6857 -6416.9249 0.2392 102.8
0.0820 -6416.6856 -6416.9178 0.2321 99.8
0.0825 -6416.6855 -6416.9084 0.2229 95.8
0.0830 -6416.6853 -6416.8987 0.2133 91.7
0.0835 -6416.6852 -6416.8909 0.2057 88.4
0.0840 -6416.6851 -6416.8875 0.2024 87.0
0.0845 -6416.6851 -6416.8897 0.2046 88.0
0.0850 -6416.6852 -6416.8978 0.2127 91.4
0.0855 -6416.6853 -6416.9104 0.2251 96.7
0.0860 -6416.6855 -6416.9250 0.2395 102.9
0.0865 -6416.6857 -6416.9386 0.2529 108.7
0.0870 -6416.6858 -6416.9487 0.2629 113.0
0.0875 -6416.6858 -6416.9538 0.2681 115.2
0.0880 -6416.6857 -6416.9543 0.2686 115.5
0.0885 -6416.6855 -6416.9515 0.2660 114.3
0.0890 -6416.6854 -6416.9478 0.2624 112.8
0.0895 -6416.6853 -6416.9449 0.2596 111.6
0.0900 -6416.6853 -6416.9440 0.2586 111.2
0.0905 -6416.6854 -6416.9448 0.2593 111.5
0.0910 -6416.6855 -6416.9460 0.2605 111.9
0.0915 -6416.6856 -6416.9459 0.2603 111.9
0.0920 -6416.6857 -6416.9428 0.2571 110.5
0.0925 -6416.6857 -6416.9358 0.2501 107.5
0.0930 -6416.6856 -6416.9248 0.2391 102.8
0.0935 -6416.6855 -6416.9109 0.2253 96.9
0.0940 -6416.6854 -6416.8960 0.2106 90.5
0.0945 -6416.6853 -6416.8824 0.1971 84.7
0.0950 -6416.6852 -6416.8721 0.1869 80.3
0.0955 -6416.6851 -6416.8666 0.1814 78.0
0.0960 -6416.6852 -6416.8660 0.1808 77.7
0.0965 -6416.6853 -6416.8696 0.1843 79.2
0.0970 -6416.6855 -6416.8754 0.1900 81.6
0.0975 -6416.6856 -6416.8814 0.1959 84.2
0.0980 -6416.6856 -6416.8859 0.2002 86.1
0.0985 -6416.6856 -6416.8879 0.2023 87.0
0.0990 -6416.6855 -6416.8880 0.2025 87.0
0.0995 -6416.6854 -6416.8875 0.2021 86.9
0.1000 -6416.6853 -6416.8884 0.2031 87.3
0.1005 -6416.6852 -6416.8920 0.2067 88.8
0.1010 -6416.6853 -6416.8989 0.2136 91.8
0.1015 -6416.6854 -6416.9086 0.2233 96.0
0.1020 -6416.6855 -6416.9197 0.2342 100.7
0.1025 -6416.6856 -6416.9303 0.2446 105.1
0.1030 -6416.6857 -6416.9386 0.2529 108.7
0.1035 -6416.6857 -6416.9437 0.2580 110.9
0.1040 -6416.6856 -6416.9452 0.2595 111.6
0.1045 -6416.6856 -6416.9438 0.2583 111.0
0.1050 -6416.6855 -6416.9409 0.2554 109.8
0.1055 -6416.6854 -6416.9377 0.2523 108.4
0.1060 -6416.6853 -6416.9357 0.2503 107.6
0.1065 -6416.6853 -6416.9356 0.2502 107.5
0.1070 -6416.6854 -6416.9372 0.2518 108.2
0.1075 -6416.6855 -6416.9398 0.2544 109.3
0.1080 -6416.6856 -6416.9420 0.2564 110.2
0.1085 -6416.6857 -6416.9421 0.2565 110.2
0.1090 -6416.6857 -6416.9392 0.2535 108.9
0.1095 -6416.6856 -6416.9328 0.2472 106.2
0.1100 -6416.6855 -6416.9239 0.2383 102.4
0.1105 -6416.6854 -6416.9139 0.2285 98.2
0.1110 -6416.6853 -6416.9045 0.2192 94.2
0.1115 -6416.6853 -6416.8972 0.2119 91.1
0.1120 -6416.6853 -6416.8923 0.2070 89.0
0.1125 -6416.6854 -6416.8896 0.2042 87.8
0.1130 -6416.6855 -6416.8879 0.2024 87.0
0.1135 -6416.6855 -6416.8860 0.2005 86.2
0.1140 -6416.6855 -6416.8830 0.1975 84.9
0.1145 -6416.6855 -6416.8785 0.1930 82.9
0.1150 -6416.6855 -6416.8728 0.1874 80.5
0.1155 -6416.6854 -6416.8668 0.1814 78.0
0.1160 -6416.6853 -6416.8617 0.1764 75.8
0.1165 -6416.6853 -6416.8587 0.1734 74.5
0.1170 -6416.6853 -6416.8585 0.1733 74.5
0.1175 -6416.6853 -6416.8616 0.1763 75.8
0.1180 -6416.6853 -6416.8674 0.1820 78.2
0.1185 -6416.6854 -6416.8749 0.1895 81.4
0.1190 -6416.6855 -6416.8829 0.1974 84.8
0.1195 -6416.6856 -6416.8901 0.2045 87.9
0.1200 -6416.6855 -6416.8958 0.2102 90.4
0.1205 -6416.6855 -6416.9001 0.2146 92.3
0.1210 -6416.6854 -6416.9040 0.2186 93.9
0.1215 -6416.6853 -6416.9085 0.2231 95.9
0.1220 -6416.6853 -6416.9147 0.2294 98.6
0.1225 -6416.6853 -6416.9230 0.2377 102.2
0.1230 -6416.6854 -6416.9331 0.2477 106.5
0.1235 -6416.6855 -6416.9441 0.2586 111.1
0.1240 -6416.6856 -6416.9547 0.2691 115.7
0.1245 -6416.6856 -6416.9637 0.2781 119.5
0.1250 -6416.6856 -6416.9707 0.2850 122.5
0.1255 -6416.6856 -6416.9753 0.2898 124.5
0.1260 -6416.6855 -6416.9782 0.2927 125.8
0.1265 -6416.6855 -6416.9801 0.2946 126.6
0.1270 -6416.6854 -6416.9817 0.2962 127.3
0.1275 -6416.6854 -6416.9836 0.2982 128.2
0.1280 -6416.6855 -6416.9861 0.3006 129.2
0.1285 -6416.6855 -6416.9886 0.3031 130.3
0.1290 -6416.6856 -6416.9905 0.3049 131.1
0.1295 -6416.6856 -6416.9905 0.3049 131.0
0.1300 -6416.6857 -6416.9875 0.3018 129.7
0.1305 -6416.6857 -6416.9807 0.2950 126.8
0.1310 -6416.6857 -6416.9699 0.2842 122.2
0.1315 -6416.6856 -6416.9559 0.2703 116.2
0.1320 -6416.6855 -6416.9400 0.2545 109.4
0.1325 -6416.6854 -6416.9240 0.2386 102.5
0.1330 -6416.6853 -6416.9092 0.2238 96.2
0.1335 -6416.6853 -6416.8964 0.2110 90.7
0.1340 -6416.6854 -6416.8856 0.2002 86.0
0.1345 -6416.6854 -6416.8763 0.1909 82.1
0.1350 -6416.6854 -6416.8680 0.1826 78.5
0.1355 -6416.6854 -6416.8602 0.1748 75.1
0.1360 -6416.6854 -6416.8531 0.1677 72.1
0.1365 -6416.6853 -6416.8470 0.1617 69.5
0.1370 -6416.6853 -6416.8428 0.1575 67.7
0.1375 -6416.6853 -6416.8413 0.1560 67.1
0.1380 -6416.6852 -6416.8430 0.1578 67.8
0.1385 -6416.6852 -6416.8484 0.1632 70.1
0.1390 -6416.6853 -6416.8573 0.1720 73.9
0.1395 -6416.6853 -6416.8691 0.1838 79.0
0.1400 -6416.6854 -6416.8831 0.1977 85.0
0.1405 -6416.6855 -6416.8981 0.2127 91.4
0.1410 -6416.6855 -6416.9131 0.2276 97.8
0.1415 -6416.6856 -6416.9271 0.2415 103.8
0.1420 -6416.6856 -6416.9394 0.2539 109.1
0.1425 -6416.6856 -6416.9500 0.2645 113.7
0.1430 -6416.6855 -6416.9592 0.2736 117.6
0.1435 -6416.6855 -6416.9673 0.2817 121.1
0.1440 -6416.6855 -6416.9744 0.2889 124.2
0.1445 -6416.6856 -6416.9804 0.2948 126.7
0.1450 -6416.6856 -6416.9845 0.2989 128.5
0.1455 -6416.6857 -6416.9859 0.3003 129.1
0.1460 -6416.6857 -6416.9841 0.2984 128.2
0.1465 -6416.6857 -6416.9788 0.2931 126.0
0.1470 -6416.6856 -6416.9705 0.2849 122.5
0.1475 -6416.6855 -6416.9602 0.2746 118.0
0.1480 -6416.6855 -6416.9488 0.2633 113.2
0.1485 -6416.6854 -6416.9375 0.2521 108.3
0.1490 -6416.6854 -6416.9269 0.2415 103.8
0.1495 -6416.6854 -6416.9175 0.2321 99.8
0.1500 -6416.6854 -6416.9093 0.2239 96.2
0.1505 -6416.6854 -6416.9021 0.2167 93.1
0.1510 -6416.6854 -6416.8960 0.2106 90.5
0.1515 -6416.6854 -6416.8908 0.2054 88.3
0.1520 -6416.6854 -6416.8869 0.2015 86.6
0.1525 -6416.6854 -6416.8843 0.1989 85.5
0.1530 -6416.6853 -6416.8834 0.1980 85.1
0.1535 -6416.6853 -6416.8846 0.1993 85.6
0.1540 -6416.6853 -6416.8881 0.2027 87.1
0.1545 -6416.6854 -6416.8936 0.2083 89.5
0.1550 -6416.6854 -6416.9007 0.2152 92.5
0.1555 -6416.6855 -6416.9081 0.2225 95.6
0.1560 -6416.6856 -6416.9146 0.2290 98.4
0.1565 -6416.6856 -6416.9194 0.2338 100.5
0.1570 -6416.6856 -6416.9218 0.2362 101.5
0.1575 -6416.6856 -6416.9221 0.2365 101.7
0.1580 -6416.6855 -6416.9208 0.2353 101.1
0.1585 -6416.6855 -6416.9189 0.2334 100.3
0.1590 -6416.6854 -6416.9171 0.2317 99.6
0.1595 -6416.6854 -6416.9160 0.2305 99.1
0.1600 -6416.6855 -6416.9154 0.2299 98.8
0.1605 -6416.6855 -6416.9149 0.2293 98.6
0.1610 -6416.6855 -6416.9139 0.2284 98.2
0.1615 -6416.6855 -6416.9122 0.2267 97.4
0.1620 -6416.6855 -6416.9099 0.2244 96.5
0.1625 -6416.6854 -6416.9076 0.2222 95.5
0.1630 -6416.6854 -6416.9062 0.2209 94.9
0.1635 -6416.6853 -6416.9067 0.2214 95.2
0.1640 -6416.6853 -6416.9099 0.2246 96.5
0.1645 -6416.6853 -6416.9161 0.2308 99.2
0.1650 -6416.6853 -6416.9253 0.2400 103.2
0.1655 -6416.6854 -6416.9369 0.2515 108.1
0.1660 -6416.6855 -6416.9496 0.2641 113.5
0.1665 -6416.6856 -6416.9619 0.2763 118.8
0.1670 -6416.6857 -6416.9724 0.2868 123.2
0.1675 -6416.6857 -6416.9800 0.2944 126.5
0.1680 -6416.6857 -6416.9844 0.2987 128.4
0.1685 -6416.6856 -6416.9856 0.3000 128.9
0.1690 -6416.6856 -6416.9844 0.2988 128.4
0.1695 -6416.6855 -6416.9814 0.2958 127.2
0.1700 -6416.6855 -6416.9772 0.2917 125.4
0.1705 -6416.6856 -6416.9720 0.2864 123.1
0.1710 -6416.6856 -6416.9652 0.2796 120.2
0.1715 -6416.6857 -6416.9562 0.2705 116.3
0.1720 -6416.6857 -6416.9441 0.2584 111.1
0.1725 -6416.6857 -6416.9287 0.2430 104.4
0.1730 -6416.6856 -6416.9102 0.2246 96.5
0.1735 -6416.6855 -6416.8897 0.2042 87.7
0.1740 -6416.6854 -6416.8686 0.1833 78.8
0.1745 -6416.6853 -6416.8490 0.1637 70.4
0.1750 -6416.6852 -6416.8326 0.1474 63.3
0.1755 -6416.6851 -6416.8208 0.1357 58.3
0.1760 -6416.6851 -6416.8146 0.1295 55.7
0.1765 -6416.6852 -6416.8138 0.1287 55.3
0.1770 -6416.6853 -6416.8176 0.1323 56.9
0.1775 -6416.6854 -6416.8244 0.1390 59.8
0.1780 -6416.6854 -6416.8330 0.1475 63.4
0.1785 -6416.6854 -6416.8423 0.1569 67.4
0.1790 -6416.6854 -6416.8522 0.1667 71.7
0.1795 -6416.6854 -6416.8628 0.1774 76.2
0.1800 -6416.6854 -6416.8747 0.1893 81.4
0.1805 -6416.6854 -6416.8884 0.2030 87.3
0.1810 -6416.6854 -6416.9039 0.2185 93.9
0.1815 -6416.6855 -6416.9206 0.2351 101.0
0.1820 -6416.6856 -6416.9371 0.2515 108.1
0.1825 -6416.6857 -6416.9517 0.2660 114.3
0.1830 -6416.6857 -6416.9630 0.2773 119.2
0.1835 -6416.6858 -6416.9700 0.2842 122.2
0.1840 -6416.6857 -6416.9723 0.2866 123.2
0.1845 -6416.6857 -6416.9706 0.2849 122.5
0.1850 -6416.6856 -6416.9656 0.2801 120.4
0.1855 -6416.6855 -6416.9587 0.2732 117.4
0.1860 -6416.6854 -6416.9510 0.2655 114.1
0.1865 -6416.6854 -6416.9435 0.2581 110.9
0.1870 -6416.6854 -6416.9368 0.2514 108.1
0.1875 -6416.6854 -6416.9312 0.2458 105.7
0.1880 -6416.6854 -6416.9264 0.2410 103.6
0.1885 -6416.6855 -6416.9221 0.2366 101.7
0.1890 -6416.6855 -6416.9178 0.2323 99.8
0.1895 -6416.6854 -6416.9135 0.2281 98.0
0.1900 -6416.6854 -6416.9099 0.2245 96.5
0.1905 -6416.6853 -6416.9074 0.2221 95.5
0.1910 -6416.6853 -6416.9070 0.2217 95.3
0.1915 -6416.6853 -6416.9090 0.2236 96.1
0.1920 -6416.6854 -6416.9132 0.2278 97.9
0.1925 -6416.6854 -6416.9189 0.2335 100.3
0.1930 -6416.6855 -6416.9248 0.2393 102.8
0.1935 -6416.6856 -6416.9296 0.2440 104.9
0.1940 -6416.6856 -6416.9321 0.2465 105.9
0.1945 -6416.6856 -6416.9320 0.2464 105.9
0.1950 -6416.6856 -6416.9294 0.2438 104.8
0.1955 -6416.6855 -6416.9250 0.2395 102.9
0.1960 -6416.6855 -6416.9197 0.2343 100.7
0.1965 -6416.6854 -6416.9142 0.2288 98.3
0.1970 -6416.6854 -6416.9091 0.2237 96.1
0.1975 -6416.6854 -6416.9046 0.2192 94.2
0.1980 -6416.6854 -6416.9007 0.2152 92.5
0.1985 -6416.6855 -6416.8970 0.2115 90.9
0.1990 -6416.6855 -6416.8933 0.2078 89.3
0.1995 -6416.6855 -6416.8892 0.2038 87.6
0.2000 -6416.6854 -6416.8850 0.1996 85.8
0.2005 -6416.6854 -6416.8813 0.1959 84.2
0.2010 -6416.6853 -6416.8789 0.1936 83.2
0.2015 -6416.6853 -6416.8791 0.1938 83.3
0.2020 -6416.6852 -6416.8826 0.1973 84.8
0.2025 -6416.6853 -6416.8897 0.2044 87.9
0.2030 -6416.6853 -6416.8999 0.2146 92.2
0.2035 -6416.6855 -6416.9120 0.2265 97.4
0.2040 -6416.6856 -6416.9242 0.2387 102.6
0.2045 -6416.6856 -6416.9350 0.2494 107.2
0.2050 -6416.6856 -6416.9434 0.2578 110.8
0.2055 -6416.6856 -6416.9492 0.2636 113.3
0.2060 -6416.6855 -6416.9528 0.2673 114.9
0.2065 -6416.6855 -6416.9553 0.2698 115.9
0.2070 -6416.6855 -6416.9573 0.2718 116.8
0.2075 -6416.6855 -6416.9593 0.2738 117.7
0.2080 -6416.6855 -6416.9612 0.2757 118.5
0.2085 -6416.6856 -6416.9624 0.2768 119.0
0.2090 -6416.6856 -6416.9619 0.2763 118.7
0.2095 -6416.6857 -6416.9589 0.2732 117.4
0.2100 -6416.6857 -6416.9528 0.2671 114.8
0.2105 -6416.6856 -6416.9436 0.2580 110.9
0.2110 -6416.6855 -6416.9322 0.2466 106.0
0.2115 -6416.6854 -6416.9198 0.2343 100.7
0.2120 -6416.6853 -6416.9081 0.2228 95.8
0.2125 -6416.6852 -6416.8990 0.2137 91.9
0.2130 -6416.6852 -6416.8934 0.2082 89.5
0.2135 -6416.6853 -6416.8917 0.2064 88.7
0.2140 -6416.6853 -6416.8931 0.2078 89.3
0.2145 -6416.6855 -6416.8963 0.2109 90.6
0.2150 -6416.6855 -6416.8996 0.2141 92.0
0.2155 -6416.6856 -6416.9016 0.2160 92.8
0.2160 -6416.6855 -6416.9018 0.2162 92.9
0.2165 -6416.6855 -6416.9008 0.2153 92.5
0.2170 -6416.6854 -6416.8999 0.2145 92.2
0.2175 -6416.6853 -6416.9008 0.2155 92.6
0.2180 -6416.6853 -6416.9045 0.2193 94.2
0.2185 -6416.6853 -6416.9115 0.2262 97.2
0.2190 -6416.6854 -6416.9209 0.2354 101.2
0.2195 -6416.6855 -6416.9312 0.2456 105.6
0.2200 -6416.6856 -6416.9405 0.2548 109.5
0.2205 -6416.6857 -6416.9469 0.2612 112.3
0.2210 -6416.6857 -6416.9494 0.2637 113.3
0.2215 -6416.6857 -6416.9476 0.2619 112.6
0.2220 -6416.6856 -6416.9423 0.2567 110.3
0.2225 -6416.6855 -6416.9350 0.2495 107.3
0.2230 -6416.6854 -6416.9277 0.2424 104.2
0.2235 -6416.6853 -6416.9222 0.2369 101.8
0.2240 -6416.6853 -6416.9196 0.2343 100.7
0.2245 -6416.6854 -6416.9198 0.2345 100.8
0.2250 -6416.6855 -6416.9220 0.2366 101.7
0.2255 -6416.6856 -6416.9245 0.2389 102.7
0.2260 -6416.6856 -6416.9254 0.2397 103.0
0.2265 -6416.6856 -6416.9236 0.2379 102.3
0.2270 -6416.6856 -6416.9191 0.2335 100.4
0.2275 -6416.6854 -6416.9132 0.2278 97.9
0.2280 -6416.6853 -6416.9079 0.2226 95.7
0.2285 -6416.6852 -6416.9054 0.2202 94.6
0.2290 -6416.6852 -6416.9070 0.2218 95.3
0.2295 -6416.6852 -6416.9130 0.2277 97.9
0.2300 -6416.6854 -6416.9224 0.2370 101.9
0.2305 -6416.6855 -6416.9333 0.2478 106.5
0.2310 -6416.6856 -6416.9436 0.2579 110.9
0.2315 -6416.6857 -6416.9513 0.2657 114.2
0.2320 -6416.6857 -6416.9555 0.2698 116.0
0.2325 -6416.6856 -6416.9560 0.2703 116.2
0.2330 -6416.6855 -6416.9537 0.2681 115.2
0.2335 -6416.6854 -6416.9502 0.2647 113.8
0.2340 -6416.6854 -6416.9472 0.2618 112.5
0.2345 -6416.6853 -6416.9461 0.2608 112.1
0.2350 -6416.6854 -6416.9473 0.2619 112.6
0.2355 -6416.6855 -6416.9500 0.2645 113.7
0.2360 -6416.6856 -6416.9526 0.2669 114.7
0.2365 -6416.6858 -6416.9526 0.2668 114.7
0.2370 -6416.6858 -6416.9479 0.2621 112.6
0.2375 -6416.6858 -6416.9374 0.2516 108.1
0.2380 -6416.6857 -6416.9214 0.2357 101.3
0.2385 -6416.6855 -6416.9018 0.2163 93.0
0.2390 -6416.6853 -6416.8814 0.1961 84.3
0.2395 -6416.6852 -6416.8634 0.1782 76.6
0.2400 -6416.6851 -6416.8501 0.1650 70.9
0.2410 -6416.6852 -6416.8397 0.1545 66.4
0.2415 -6416.6854 -6416.8407 0.1553 66.7
0.2420 -6416.6855 -6416.8431 0.1577 67.8
0.2425 -6416.6855 -6416.8455 0.1599 68.7
0.2430 -6416.6855 -6416.8468 0.1613 69.3
0.2435 -6416.6854 -6416.8474 0.1620 69.6
0.2440 -6416.6853 -6416.8486 0.1633 70.2
0.2445 -6416.6852 -6416.8521 0.1669 71.7
0.2450 -6416.6852 -6416.8595 0.1744 74.9
0.2455 -6416.6852 -6416.8719 0.1867 80.2
0.2460 -6416.6853 -6416.8889 0.2036 87.5
0.2465 -6416.6854 -6416.9091 0.2237 96.1
0.2470 -6416.6856 -6416.9299 0.2442 105.0
0.2475 -6416.6858 -6416.9482 0.2624 112.8
0.2480 -6416.6859 -6416.9616 0.2757 118.5
0.2485 -6416.6859 -6416.9686 0.2828 121.5
0.2490 -6416.6857 -6416.9696 0.2838 122.0
0.2495 -6416.6856 -6416.9661 0.2805 120.6
0.2500 -6416.6854 -6416.9608 0.2754 118.4
True
systems = dpdata.LabeledSystem("md.traj", fmt="ase/traj")
preview_stride = max(1, MD_STEPS // 10)
for frame_idx in range(0, len(systems), preview_stride):
print("frame", frame_idx)
systems[frame_idx].to_3dmol(
f_idx=0,
size=(450, 350),
).show()
frame 0
3Dmol.js failed to load for some reason. Please check your browser console for error messages.
frame 50
3Dmol.js failed to load for some reason. Please check your browser console for error messages.
frame 100
3Dmol.js failed to load for some reason. Please check your browser console for error messages.
frame 150
3Dmol.js failed to load for some reason. Please check your browser console for error messages.
frame 200
3Dmol.js failed to load for some reason. Please check your browser console for error messages.
frame 250
3Dmol.js failed to load for some reason. Please check your browser console for error messages.
frame 300
3Dmol.js failed to load for some reason. Please check your browser console for error messages.
frame 350
3Dmol.js failed to load for some reason. Please check your browser console for error messages.
frame 400
3Dmol.js failed to load for some reason. Please check your browser console for error messages.
frame 450
3Dmol.js failed to load for some reason. Please check your browser console for error messages.
frame 500
3Dmol.js failed to load for some reason. Please check your browser console for error messages.
3.4.3. Save the trajectory as a DeePMD dataset#
The dpdata package writes the Teacher-labeled trajectory into a DeePMD-kit data directory. The energy and force labels come from the DPA3 Teacher model used during MD. Virial labels are not used in this lightweight documentation example, so the Student training configuration sets the virial loss weights to zero.
Reference: dpdata integration
systems = dpdata.LabeledSystem("md.traj", fmt="ase/traj")
systems.to_deepmd_npy("data/cyclohexane/train/system_001")
!tree data/cyclohexane/train/system_001
data/cyclohexane/train/system_001
├── nopbc
├── set.000
│ ├── box.npy
│ ├── coord.npy
│ ├── energy.npy
│ └── force.npy
├── type.raw
└── type_map.raw
2 directories, 7 files
3.5. Generate the Student training configuration#
The DeePMD-kit training entry point is a JSON configuration file. For clarity, this tutorial writes the configuration to configs/student_se_atten_v2.json instead of placing it directly inside the run directory.
If you want to adjust model size, the number of training steps, learning-rate decay, or loss weights, start from the variables and JSON fields in the following cell.
The Student model uses the se_atten_v2 descriptor and learns the DPA-3 Teacher energy and force labels for the sampled cyclohexane configurations.
%%bash
set -euo pipefail
mkdir -p configs runs/cyclohexane_dpa3_distill/student_se_atten_v2
cat > configs/student_se_atten_v2.json <<EOF
{
"model": {
"type_map": ["C", "H"],
"descriptor": {
"type": "se_atten_v2",
"sel": "auto",
"rcut_smth": 0.5,
"rcut": 6.0,
"neuron": [25, 50, 100],
"resnet_dt": false,
"axis_neuron": 16,
"seed": 1,
"attn": 128,
"attn_layer": 0,
"attn_dotr": true,
"attn_mask": false,
"precision": "float32"
},
"fitting_net": {
"neuron": [240, 240, 240],
"resnet_dt": true,
"precision": "float32",
"seed": 1
}
},
"learning_rate": {
"type": "exp",
"decay_steps": 10000,
"start_lr": 1e-3,
"stop_lr": 1e-5
},
"loss": {
"type": "ener",
"start_pref_e": 0.02,
"limit_pref_e": 1.0,
"start_pref_f": 1000,
"limit_pref_f": 1.0,
"start_pref_v": 0,
"limit_pref_v": 0
},
"training": {
"training_data": {
"systems": ["$(pwd)/data/cyclohexane/train/system_001"],
"batch_size": 8
},
"numb_steps": 20000,
"seed": 10,
"disp_file": "lcurve.out",
"disp_freq": 200,
"save_freq": 5000,
"save_ckpt": "model.ckpt",
"disp_training": true,
"time_training": true,
"profiling": false,
"profiling_file": "timeline.json"
}
}
EOF
3.6. Train the Student model#
This cell starts DeePMD-kit training.
The input configuration is configs/student_se_atten_v2.json. The training outputs are written to runs/cyclohexane_dpa3_distill/student_se_atten_v2.
%%bash
set -euo pipefail
STUDENT_INPUT="$(pwd)/configs/student_se_atten_v2.json"
STUDENT_WORKDIR="$(pwd)/runs/cyclohexane_dpa3_distill/student_se_atten_v2"
mkdir -p "$STUDENT_WORKDIR"
cd "$STUDENT_WORKDIR"
dp --pt train "$STUDENT_INPUT"
To get the best performance, it is recommended to adjust the number of threads by setting the environment variables OMP_NUM_THREADS, DP_INTRA_OP_PARALLELISM_THREADS, and DP_INTER_OP_PARALLELISM_THREADS. See https://deepmd.rtfd.io/parallelism/ for more information.
[2026-06-20 22:11:19,616] DEEPMD INFO DeePMD version: 3.1.3
[2026-06-20 22:11:19,617] DEEPMD INFO Configuration path: //configs/student_se_atten_v2.json
[2026-06-20 22:11:19,618] DEEPMD INFO _____ _____ __ __ _____ _ _ _
[2026-06-20 22:11:19,619] DEEPMD INFO | __ \ | __ \ | \/ || __ \ | | (_)| |
[2026-06-20 22:11:19,619] DEEPMD INFO | | | | ___ ___ | |__) || \ / || | | | ______ | | __ _ | |_
[2026-06-20 22:11:19,619] DEEPMD INFO | | | | / _ \ / _ \| ___/ | |\/| || | | ||______|| |/ /| || __|
[2026-06-20 22:11:19,619] DEEPMD INFO | |__| || __/| __/| | | | | || |__| | | < | || |_
[2026-06-20 22:11:19,619] DEEPMD INFO |_____/ \___| \___||_| |_| |_||_____/ |_|\_\|_| \__|
[2026-06-20 22:11:19,619] DEEPMD INFO Please read and cite:
[2026-06-20 22:11:19,619] DEEPMD INFO Wang, Zhang, Han and E, Comput.Phys.Comm. 228, 178-184 (2018)
[2026-06-20 22:11:19,619] DEEPMD INFO Zeng et al, J. Chem. Phys., 159, 054801 (2023)
[2026-06-20 22:11:19,619] DEEPMD INFO Zeng et al, J. Chem. Theory Comput., 21, 4375-4385 (2025)
[2026-06-20 22:11:19,619] DEEPMD INFO See https://deepmd.rtfd.io/credits/ for details.
[2026-06-20 22:11:19,619] DEEPMD INFO ---------------------------------------------------------------------------------------------------------------
[2026-06-20 22:11:19,619] DEEPMD INFO Installed to: /opt/mamba/lib/python3.12/site-packages/deepmd
[2026-06-20 22:11:19,619] DEEPMD INFO Source: v3.1.3
[2026-06-20 22:11:19,619] DEEPMD INFO Source Branch: HEAD
[2026-06-20 22:11:19,619] DEEPMD INFO Source Commit: b2c8511e
[2026-06-20 22:11:19,619] DEEPMD INFO Source Commit at: 2026-03-18 14:20:41 +0000
[2026-06-20 22:11:19,619] DEEPMD INFO Float Precision: Double
[2026-06-20 22:11:19,619] DEEPMD INFO Build Variant: CUDA
[2026-06-20 22:11:19,619] DEEPMD INFO Backend: PyTorch
[2026-06-20 22:11:19,619] DEEPMD INFO PT Ver: v2.10.0+cu128-g449b1768410
[2026-06-20 22:11:19,619] DEEPMD INFO Custom OP Enabled: True
[2026-06-20 22:11:19,620] DEEPMD INFO Built with PT Ver: 2.10.0
[2026-06-20 22:11:19,620] DEEPMD INFO Built with PT Inc: /tmp/build-env-wnfrgr6r/lib/python3.11/site-packages/torch/include
[2026-06-20 22:11:19,620] DEEPMD INFO /tmp/build-env-wnfrgr6r/lib/python3.11/site-packages/torch/include/torch/csrc/api/include
[2026-06-20 22:11:19,620] DEEPMD INFO Built with PT Lib: /tmp/build-env-wnfrgr6r/lib/python3.11/site-packages/torch/lib
[2026-06-20 22:11:19,620] DEEPMD INFO Running on: bohrium-156-1473861
[2026-06-20 22:11:19,620] DEEPMD INFO Computing Device: CPU
[2026-06-20 22:11:19,620] DEEPMD INFO CUDA_VISIBLE_DEVICES: unset
[2026-06-20 22:11:19,620] DEEPMD INFO Visible GPU Count: 0
[2026-06-20 22:11:19,620] DEEPMD INFO Num Intra Threads: 0
[2026-06-20 22:11:19,620] DEEPMD INFO Num Inter Threads: 0
[2026-06-20 22:11:19,620] DEEPMD INFO ---------------------------------------------------------------------------------------------------------------
[2026-06-20 22:11:19,708] DEEPMD INFO Calculate neighbor statistics... (add --skip-neighbor-stat to skip this step)
[2026-06-20 22:11:19,727] DEEPMD WARNING You can use the environment variable DP_INFER_BATCH_SIZE tocontrol the inference batch size (nframes * natoms). The default value is 1024.
[2026-06-20 22:11:19,864] DEEPMD INFO Neighbor statistics: training data with minimal neighbor distance: 1.067591
[2026-06-20 22:11:19,864] DEEPMD INFO Neighbor statistics: training data with maximum neighbor size: [17] (cutoff radius: 6.000000)
[2026-06-20 22:11:19,877] DEEPMD INFO Constructing DataLoaders from 1 systems
[2026-06-20 22:11:19,943] DEEPMD INFO Packing data for statistics from 1 systems
[2026-06-20 22:11:20,134] DEEPMD INFO RMSE of energy per atom after linear regression is: 0.0023592935450802953 in the unit of energy.
[2026-06-20 22:11:21,910] DEEPMD INFO ---Summary of DataSystem: Training -----------------------------------------------
[2026-06-20 22:11:21,911] DEEPMD INFO Found 1 System(s):
[2026-06-20 22:11:21,911] DEEPMD INFO system natoms bch_sz n_bch prob pbc
[2026-06-20 22:11:21,911] DEEPMD INFO //data/cyclohexane/train/system_001 18 8 62 1.000e+00 F
[2026-06-20 22:11:21,911] DEEPMD INFO --------------------------------------------------------------------------------------
[2026-06-20 22:11:21,914] DEEPMD INFO Model Params: 0.516 M (Trainable: 0.516 M)
[2026-06-20 22:11:21,914] DEEPMD INFO Start to train 20000 steps.
/opt/mamba/lib/python3.12/site-packages/torch/utils/data/dataloader.py:1118: UserWarning: 'pin_memory' argument is set as true but no accelerator is found, then device pinned memory won't be used.
super().__init__(loader)
[2026-06-20 22:11:23,219] DEEPMD INFO Batch 1: trn: rmse = 5.05e+01, rmse_e = 1.41e+00, rmse_f = 1.60e+00, lr = 1.00e-03
[2026-06-20 22:11:23,224] DEEPMD INFO Batch 1: total wall time = 1.31 s, eta = 7:15:57
[2026-06-20 22:11:43,185] DEEPMD INFO Batch 200: trn: rmse = 1.06e+01, rmse_e = 1.55e+00, rmse_f = 3.35e-01, lr = 1.00e-03
[2026-06-20 22:11:43,191] DEEPMD INFO Batch 200: total wall time = 19.97 s, eta = 0:32:56
[2026-06-20 22:12:03,518] DEEPMD INFO Batch 400: trn: rmse = 1.21e+01, rmse_e = 7.73e-01, rmse_f = 3.81e-01, lr = 1.00e-03
[2026-06-20 22:12:03,523] DEEPMD INFO Batch 400: total wall time = 20.33 s, eta = 0:33:12
[2026-06-20 22:12:23,020] DEEPMD INFO Batch 600: trn: rmse = 1.09e+01, rmse_e = 3.15e-01, rmse_f = 3.44e-01, lr = 1.00e-03
[2026-06-20 22:12:23,026] DEEPMD INFO Batch 600: total wall time = 19.50 s, eta = 0:31:31
[2026-06-20 22:12:42,687] DEEPMD INFO Batch 800: trn: rmse = 1.02e+01, rmse_e = 1.57e-01, rmse_f = 3.21e-01, lr = 1.00e-03
[2026-06-20 22:12:42,696] DEEPMD INFO Batch 800: total wall time = 19.67 s, eta = 0:31:28
[2026-06-20 22:13:01,905] DEEPMD INFO Batch 1000: trn: rmse = 9.18e+00, rmse_e = 3.84e-02, rmse_f = 2.90e-01, lr = 1.00e-03
[2026-06-20 22:13:01,913] DEEPMD INFO Batch 1000: total wall time = 19.22 s, eta = 0:30:25
[2026-06-20 22:13:21,704] DEEPMD INFO Batch 1200: trn: rmse = 9.27e+00, rmse_e = 2.77e-02, rmse_f = 2.93e-01, lr = 1.00e-03
[2026-06-20 22:13:21,711] DEEPMD INFO Batch 1200: total wall time = 19.80 s, eta = 0:31:01
[2026-06-20 22:13:41,121] DEEPMD INFO Batch 1400: trn: rmse = 7.77e+00, rmse_e = 1.45e-01, rmse_f = 2.46e-01, lr = 1.00e-03
[2026-06-20 22:13:41,126] DEEPMD INFO Batch 1400: total wall time = 19.41 s, eta = 0:30:05
[2026-06-20 22:14:00,631] DEEPMD INFO Batch 1600: trn: rmse = 7.81e+00, rmse_e = 1.35e-01, rmse_f = 2.47e-01, lr = 1.00e-03
[2026-06-20 22:14:00,639] DEEPMD INFO Batch 1600: total wall time = 19.51 s, eta = 0:29:55
[2026-06-20 22:14:20,736] DEEPMD INFO Batch 1800: trn: rmse = 7.61e+00, rmse_e = 1.10e-01, rmse_f = 2.41e-01, lr = 1.00e-03
[2026-06-20 22:14:20,742] DEEPMD INFO Batch 1800: total wall time = 20.10 s, eta = 0:30:29
[2026-06-20 22:14:40,227] DEEPMD INFO Batch 2000: trn: rmse = 8.04e+00, rmse_e = 1.45e-01, rmse_f = 2.54e-01, lr = 1.00e-03
[2026-06-20 22:14:40,238] DEEPMD INFO Batch 2000: total wall time = 19.50 s, eta = 0:29:14
[2026-06-20 22:14:59,830] DEEPMD INFO Batch 2200: trn: rmse = 8.64e+00, rmse_e = 1.03e-01, rmse_f = 2.73e-01, lr = 1.00e-03
[2026-06-20 22:14:59,836] DEEPMD INFO Batch 2200: total wall time = 19.60 s, eta = 0:29:04
[2026-06-20 22:15:20,064] DEEPMD INFO Batch 2400: trn: rmse = 7.19e+00, rmse_e = 8.45e-02, rmse_f = 2.28e-01, lr = 1.00e-03
[2026-06-20 22:15:20,069] DEEPMD INFO Batch 2400: total wall time = 20.23 s, eta = 0:29:40
[2026-06-20 22:15:39,505] DEEPMD INFO Batch 2600: trn: rmse = 6.55e+00, rmse_e = 5.15e-02, rmse_f = 2.07e-01, lr = 1.00e-03
[2026-06-20 22:15:39,512] DEEPMD INFO Batch 2600: total wall time = 19.44 s, eta = 0:28:11
[2026-06-20 22:15:59,096] DEEPMD INFO Batch 2800: trn: rmse = 5.76e+00, rmse_e = 1.31e-03, rmse_f = 1.82e-01, lr = 1.00e-03
[2026-06-20 22:15:59,101] DEEPMD INFO Batch 2800: total wall time = 19.59 s, eta = 0:28:04
[2026-06-20 22:16:18,786] DEEPMD INFO Batch 3000: trn: rmse = 5.16e+00, rmse_e = 8.86e-03, rmse_f = 1.63e-01, lr = 1.00e-03
[2026-06-20 22:16:18,793] DEEPMD INFO Batch 3000: total wall time = 19.69 s, eta = 0:27:53
[2026-06-20 22:16:38,439] DEEPMD INFO Batch 3200: trn: rmse = 4.74e+00, rmse_e = 3.30e-02, rms[2026-06-20 22:16:58,093] DEEPMD INFO Batch 3400: trn: rmse = 5.01e+00, rmse_e = 8.76e-02, rmse_f = 1.58e-01, lr = 1.00e-03
[2026-06-20 22:16:58,103] DEEPMD INFO Batch 3400: total wall time = 19.66 s, eta = 0:27:11
[2026-06-20 22:17:18,219] DEEPMD INFO Batch 3600: trn: rmse = 7.31e+00, rmse_e = 2.97e-02, rmse_f = 2.31e-01, lr = 1.00e-03
[2026-06-20 22:17:18,231] DEEPMD INFO Batch 3600: total wall time = 20.13 s, eta = 0:27:30
[2026-06-20 22:17:38,310] DEEPMD INFO Batch 3800: trn: rmse = 4.72e+00, rmse_e = 1.95e-01, rmse_f = 1.49e-01, lr = 1.00e-03
[2026-06-20 22:17:38,317] DEEPMD INFO Batch 3800: total wall time = 20.09 s, eta = 0:27:06
[2026-06-20 22:17:57,832] DEEPMD INFO Batch 4000: trn: rmse = 4.76e+00, rmse_e = 3.94e-02, rmse_f = 1.51e-01, lr = 1.00e-03
[2026-06-20 22:17:57,841] DEEPMD INFO Batch 4000: total wall time = 19.52 s, eta = 0:26:01
[2026-06-20 22:18:17,989] DEEPMD INFO Batch 4200: trn: rmse = 4.53e+00, rmse_e = 3.34e-02, rmse_f = 1.43e-01, lr = 1.00e-03
[2026-06-20 22:18:17,995] DEEPMD INFO Batch 4200: total wall time = 20.16 s, eta = 0:26:32
[2026-06-20 22:18:37,832] DEEPMD INFO Batch 4400: trn: rmse = 4.07e+00, rmse_e = 1.05e-02, rmse_f = 1.29e-01, lr = 1.00e-03
[2026-06-20 22:18:37,837] DEEPMD INFO Batch 4400: total wall time = 19.84 s, eta = 0:25:47
[2026-06-20 22:18:58,006] DEEPMD INFO Batch 4600: trn: rmse = 3.88e+00, rmse_e = 8.25e-03, rmse_f = 1.23e-01, lr = 1.00e-03
[2026-06-20 22:18:58,016] DEEPMD INFO Batch 4600: total wall time = 20.18 s, eta = 0:25:53
[2026-06-20 22:19:17,887] DEEPMD INFO Batch 4800: trn: rmse = 5.34e+00, rmse_e = 7.78e-03, rmse_f = 1.69e-01, lr = 1.00e-03
[2026-06-20 22:19:17,892] DEEPMD INFO Batch 4800: total wall time = 19.87 s, eta = 0:25:10
[2026-06-20 22:19:37,477] DEEPMD INFO Batch 5000: trn: rmse = 3.65e+00, rmse_e = 5.98e-02, rmse_f = 1.15e-01, lr = 1.00e-03
[2026-06-20 22:19:37,485] DEEPMD INFO Batch 5000: total wall time = 19.59 s, eta = 0:24:29
[2026-06-20 22:19:37,563] DEEPMD INFO Saved model to model.ckpt-5000.pt
[2026-06-20 22:19:57,347] DEEPMD INFO Batch 5200: trn: rmse = 3.78e+00, rmse_e = 3.19e-02, rmse_f = 1.19e-01, lr = 1.00e-03
[2026-06-20 22:19:57,354] DEEPMD INFO Batch 5200: total wall time = 19.87 s, eta = 0:24:30
[2026-06-20 22:20:17,420] DEEPMD INFO Batch 5400: trn: rmse = 3.89e+00, rmse_e = 7.17e-02, rmse_f = 1.23e-01, lr = 1.00e-03
[2026-06-20 22:20:17,428] DEEPMD INFO Batch 5400: total wall time = 20.07 s, eta = 0:24:25
[2026-06-20 22:20:36,939] DEEPMD INFO Batch 5600: trn: rmse = 4.47e+00, rmse_e = 8.18e-02, rmse_f = 1.41e-01, lr = 1.00e-03
[2026-06-20 22:20:36,945] DEEPMD INFO Batch 5600: total wall time = 19.52 s, eta = 0:23:25
[2026-06-20 22:20:56,608] DEEPMD INFO Batch 5800: trn: rmse = 3.40e+00, rmse_e = 3.34e-02, rmse_f = 1.07e-01, lr = 1.00e-03
[2026-06-20 22:20:56,612] DEEPMD INFO Batch 5800: total wall time = 19.67 s, eta = 0:23:16
[2026-06-20 22:21:16,699] DEEPMD INFO Batch 6000: trn: rmse = 3.32e+00, rmse_e = 2.66e-02, rmse_f = 1.05e-01, lr = 1.00e-03
[2026-06-20 22:21:16,706] DEEPMD INFO Batch 6000: total wall time = 20.09 s, eta = 0:23:26
[2026-06-20 22:21:36,669] DEEPMD INFO Batch 6200: trn: rmse = 2.52e+00, rmse_e = 4.84e-02, rmse_f = 7.98e-02, lr = 1.00e-03
[2026-06-20 22:21:36,676] DEEPMD INFO Batch 6200: total wall time = 19.97 s, eta = 0:22:57
[2026-06-20 22:21:56,198] DEEPMD INFO Batch 6400: trn: rmse = 2.92e+00, rmse_e = 1.34e-01, rmse_f = 9.24e-02, lr = 1.00e-03
[2026-06-20 22:21:56,205] DEEPMD INFO Batch 6400: total wall time = 19.53 s, eta = 0:22:07
[2026-06-20 22:22:16,126] DEEPMD INFO Batch 6600: trn: rmse = 2.72e+00, rmse_e = 1.08e-01, rmse_f = 8.60e-02, lr = 1.00e-03
[2026-06-20 22:22:16,130] DEEPMD INFO Batch 6600: total wall time = 19.92 s, eta = 0:22:14
[2026-06-20 22:22:35,509] DEEPMD INFO Batch 6800: trn: rmse = 2.62e+00, rmse_e = 9.56e-02, rmse_f = 8.28e-02, lr = 1.00e-03
[2026-06-20 22:22:35,516] DEEPMD INFO Batch 6800: total wall time = 19.39 s, eta = 0:21:19
[2026-06-20 22:22:55,250] DEEPMD INFO Batch 7000: trn: rmse = 2.32e+00, rmse_e = 2.76e-03, rmse_f = 7.35e-02, lr = 1.00e-03
[2026-06-20 22:22:55,258] DEEPMD INFO Batch 7000: total wall time = 19.74 s, eta = 0:21:23
[2026-06-20 22:23:15,219] DEEPMD INFO Batch 7200: trn: rmse = 2.89e+00, rmse_e = 2.74e-02, rmse_f = 9.15e-02, lr = 1.00e-03
[2026-06-20 22:23:15,223] DEEPMD INFO Batch 7200: total wall time = 19.97 s, eta = 0:21:17
[2026-06-20 22:23:34,817] DEEPMD INFO Batch 7400: trn: rmse = 2.12e+00, rmse_e = 1.83e-03, rmse_f = 6.72e-02, lr = 1.00e-03
[2026-06-20 22:23:34,827] DEEPMD INFO Batch 7400: total wall time = 19.60 s, eta = 0:20:34
[2026-06-20 22:23:54,562] DEEPMD INFO Batch 7600: trn: rmse = 2.03e+00, rmse_e = 6.56e-02, rmse_f = 6.43e-02, lr = 1.00e-03
[2026-06-20 22:23:54,566] DEEPMD INFO Batch 7600: total wall time = 19.74 s, eta = 0:20:23
[2026-06-20 22:24:14,455] DEEPMD INFO Batch 7800: trn: rmse = 2.75e+00, rmse_e = 2.32e-02, rmse_f = 8.70e-02, lr = 1.00e-03
[2026-06-20 22:24:14,462] DEEPMD INFO Batch 7800: total wall time = 19.90 s, eta = 0:20:13
[2026-06-20 22:24:33,934] DEEPMD INFO Batch 8000: trn: rmse = 2.16e+00, rmse_e = 5.83e-03, rmse_f = 6.83e-02, lr = 1.00e-03
[2026-06-20 22:24:33,941] DEEPMD INFO Batch 8000: total wall time = 19.48 s, eta = 0:19:28
[2026-06-20 22:24:53,671] DEEPMD INFO Batch 8200: trn: rmse = 1.92e+00, rmse_e = 2.18e-02, rmse_f = 6.09e-02, lr = 1.00e-03
[2026-06-20 22:24:53,676] DEEPMD INFO Batch 8200: total wall time = 19.73 s, eta = 0:19:24
[2026-06-20 22:25:13,490] DEEPMD INFO Batch 8400: trn: rmse = 1.98e+00, rmse_e = 1.23e-02, rmse_f = 6.26e-02, lr = 1.00e-03
[2026-06-20 22:25:13,497] DEEPMD INFO Batch 8400: total wall time = 19.82 s, eta = 0:19:09
[2026-06-20 22:25:32,778] DEEPMD INFO Batch 8600: trn: rmse = 2.14e+00, rmse_e = 2.84e-02, rmse_f = 6.76e-02, lr = 1.00e-03
[2026-06-20 22:25:32,782] DEEPMD INFO Batch 8600: total wall time = 19.28 s, eta = 0:18:19
[2026-06-20 22:25:52,329] DEEPMD INFO Batch 8800: trn: rmse = 2.11e+00, rmse_e = 1.83e-03, rmse_f = 6.67e-02, lr = 1.00e-03
[2026-06-20 22:25:52,336] DEEPMD INFO Batch 8800: total wall time = 19.55 s, eta = 0:18:14
[2026-06-20 22:26:12,138] DEEPMD INFO Batch 9000: trn: rmse = 1.96e+00, rmse_e = 5.44e-02, rmse_f = 6.19e-02, lr = 1.00e-03
[2026-06-20 22:26:12,145] DEEPMD INFO Batch 9000: total wall time = 19.81 s, eta = 0:18:09
[2026-06-20 22:26:31,658] DEEPMD INFO Batch 9200: trn: rmse = 2.00e+00, rmse_e = 6.88e-03, rmse_f = 6.33e-02, lr = 1.00e-03
[2026-06-20 22:26:31,663] DEEPMD INFO Batch 9200: total wall time = 19.52 s, eta = 0:17:33
[2026-06-20 22:26:51,327] DEEPMD INFO Batch 9400: trn: rmse = 2.09e+00, rmse_e = 2.45e-02, rmse_f = 6.62e-02, lr = 1.00e-03
[2026-06-20 22:26:51,334] DEEPMD INFO Batch 9400: total wall time = 19.67 s, eta = 0:17:22
[2026-06-20 22:27:11,550] DEEPMD INFO Batch 9600: trn: rmse = 2.46e+00, rmse_e = 2.66e-02, rmse_f = 7.78e-02, lr = 1.00e-03
[2026-06-20 22:27:11,555] DEEPMD INFO Batch 9600: total wall time = 20.22 s, eta = 0:17:31
[2026-06-20 22:27:30,951] DEEPMD INFO Batch 9800: trn: rmse = 1.94e+00, rmse_e = 3.19e-02, rmse_f = 6.13e-02, lr = 1.00e-03
[2026-06-20 22:27:30,959] DEEPMD INFO Batch 9800: total wall time = 19.40 s, eta = 0:16:29
[2026-06-20 22:27:50,736] DEEPMD INFO Batch 10000: trn: rmse = 1.63e+00, rmse_e = 1.29e-02, rmse_f = 5.15e-02, lr = 1.00e-03
[2026-06-20 22:27:50,743] DEEPMD INFO Batch 10000: total wall time = 19.78 s, eta = 0:16:29
[2026-06-20 22:27:50,805] DEEPMD INFO Saved model to model.ckpt-10000.pt
[2026-06-20 22:28:10,455] DEEPMD INFO Batch 10200: trn: rmse = 4.94e-01, rmse_e = 3.15e-04, rmse_f = 4.91e-02, lr = 1.00e-04
[2026-06-20 22:28:10,461] DEEPMD INFO Batch 10200: total wall time = 19.72 s, eta = 0:16:06
[2026-06-20 22:28:29,744] DEEPMD INFO Batch 10400: trn: rmse = 5.30e-01, rmse_e = 5.29e-04, rmse_f = 5.27e-02, lr = 1.00e-04
[2026-06-20 22:28:29,751] DEEPMD INFO Batch 10400: total wall time = 19.29 s, eta = 0:15:25
[2026-06-20 22:28:49,551] DEEPMD INFO Batch 10600: trn: rmse = 5.47e-01, rmse_e = 4.92e-04, rmse_f = 5.45e-02, lr = 1.00e-04
[2026-06-20 22:28:49,555] DEEPMD INFO Batch 10600: total wall time = 19.80 s, eta = 0:15:30
[2026-06-20 22:29:09,307] DEEPMD INFO Batch 10800: trn: rmse = 4.89e-01, rmse_e = 2.95e-04, rmse_f = 4.87e-02, lr = 1.00e-04
[2026-06-20 22:29:09,308] DEEPMD INFO Batch 10800: total wall time = 19.75 s, eta = 0:15:08
[2026-06-20 22:29:28,803] DEEPMD INFO Batch 11000: trn: rmse = 4.98e-01, rmse_e = 4.36e-04, rmse_f = 4.95e-02, lr = 1.00e-04
[2026-06-20 22:29:28,809] DEEPMD INFO Batch 11000: total wall time = 19.50 s, eta = 0:14:37
[2026-06-20 22:29:48,380] DEEPMD INFO Batch 11200: trn: rmse = 4.90e-01, rmse_e = 4.00e-04, rmse_f = 4.87e-02, lr = 1.00e-04
[2026-06-20 22:29:48,386] DEEPMD INFO Batch 11200: total wall time = 19.58 s, eta = 0:14:21
[2026-06-20 22:30:08,461] DEEPMD INFO Batch 11400: trn: rmse = 5.06e-01, rmse_e = 4.52e-04, rmse_f = 5.04e-02, lr = 1.00e-04
[2026-06-20 22:30:08,467] DEEPMD INFO Batch 11400: total wall time = 20.08 s, eta = 0:14:23
[2026-06-20 22:30:28,330] DEEPMD INFO Batch 11600: trn: rmse = 5.23e-01, rmse_e = 5.25e-04, rmse_f = 5.21e-02, lr = 1.00e-04
[2026-06-20 22:30:28,335] DEEPMD INFO Batch 11600: total wall time = 19.87 s, eta = 0:13:54
[2026-06-20 22:30:47,928] DEEPMD INFO Batch 11800: trn: rmse = 4.26e-01, rmse_e = 4.76e-04, rmse_f = 4.24e-02, lr = 1.00e-04
[2026-06-20 22:30:47,937] DEEPMD INFO Batch 11800: total wall time = 19.60 s, eta = 0:13:23
[2026-06-20 22:31:07,697] DEEPMD INFO Batch 12000: trn: rmse = 4.88e-01, rmse_e = 5.28e-04, rmse_f = 4.85e-02, lr = 1.00e-04
[2026-06-20 22:31:07,703] DEEPMD INFO Batch 12000: total wall time = 19.77 s, eta = 0:13:10
[2026-06-20 22:31:26,948] DEEPMD INFO Batch 12200: trn: rmse = 4.46e-01, rmse_e = 5.08e-04, rmse_f = 4.44e-02, lr = 1.00e-04
[2026-06-20 22:31:26,956] DEEPMD INFO Batch 12200: total wall time = 19.25 s, eta = 0:12:30
[2026-06-20 22:31:46,819] DEEPMD INFO Batch 12400: trn: rmse = 5.06e-01, rmse_e = 4.20e-04, rmse_f = 5.04e-02, lr = 1.00e-04
[2026-06-20 22:31:46,825] DEEPMD INFO Batch 12400: total wall time = 19.87 s, eta = 0:12:35
[2026-06-20 22:32:06,540] DEEPMD INFO Batch 12600: trn: rmse = 5.13e-01, rmse_e = 4.20e-04, rmse_f = 5.11e-02, lr = 1.00e-04
[2026-06-20 22:32:06,545] DEEPMD INFO Batch 12600: total wall time = 19.72 s, eta = 0:12:09
[2026-06-20 22:32:26,238] DEEPMD INFO Batch 12800: trn: rmse = 4.72e-01, rmse_e = 6.15e-04, rmse_f = 4.70e-02, lr = 1.00e-04
[2026-06-20 22:32:26,245] DEEPMD INFO Batch 12800: total wall time = 19.70 s, eta = 0:11:49
[2026-06-20 22:32:46,038] DEEPMD INFO Batch 13000: trn: rmse = 4.52e-01, rmse_e = 3.20e-04, rmse_f = 4.50e-02, lr = 1.00e-04
[2026-06-20 22:32:46,043] DEEPMD INFO Batch 13000: total wall time = 19.80 s, eta = 0:11:32
[2026-06-20 22:33:05,785] DEEPMD INFO Batch 13200: trn: rmse = 4.90e-01, rmse_e = 3.93e-04, rmse_f = 4.88e-02, lr = 1.00e-04
[2026-06-20 22:33:05,790] DEEPMD INFO Batch 13200: total wall time = 19.75 s, eta = 0:11:11
[2026-06-20 22:33:25,338] DEEPMD INFO Batch 13400: trn: rmse = 4.96e-01, rmse_e = 7.06e-04, rmse_f = 4.93e-02, lr = 1.00e-04
[2026-06-20 22:33:25,342] DEEPMD INFO Batch 13400: total wall time = 19.55 s, eta = 0:10:45
[2026-06-20 22:33:45,293] DEEPMD INFO Batch 13600: trn: rmse = 4.18e-01, rmse_e = 4.06e-04, rmse_f = 4.16e-02, lr = 1.00e-04
[2026-06-20 22:33:45,300] DEEPMD INFO Batch 13600: total wall time = 19.96 s, eta = 0:10:38
[2026-06-20 22:34:04,836] DEEPMD INFO Batch 13800: trn: rmse = 4.95e-01, rmse_e = 4.57e-04, rmse_f = 4.93e-02, lr = 1.00e-04
[2026-06-20 22:34:04,842] DEEPMD INFO Batch 13800: total wall time = 19.54 s, eta = 0:10:05
[2026-06-20 22:34:24,322] DEEPMD INFO Batch 14000: trn: rmse = 4.81e-01, rmse_e = 5.96e-04, rmse_f = 4.79e-02, lr = 1.00e-04
[2026-06-20 22:34:24,327] DEEPMD INFO Batch 14000: total wall time = 19.49 s, eta = 0:09:44
[2026-06-20 22:34:43,839] DEEPMD INFO Batch 14200: trn: rmse = 4.78e-01, rmse_e = 1.45e-03, rmse_f = 4.75e-02, lr = 1.00e-04
[2026-06-20 22:34:43,846] DEEPMD INFO Batch 14200: total wall time = 19.52 s, eta = 0:09:26
[2026-06-20 22:35:03,654] DEEPMD INFO Batch 14400: trn: rmse = 3.88e-01, rmse_e = 1.10e-03, rmse_f = 3.86e-02, lr = 1.00e-04
[2026-06-20 22:35:03,660] DEEPMD INFO Batch 14400: total wall time = 19.81 s, eta = 0:09:14
[2026-06-20 22:35:23,171] DEEPMD INFO Batch 14600: trn: rmse = 5.43e-01, rmse_e = 1.30e-03, rmse_f = 5.40e-02, lr = 1.00e-04
[2026-06-20 22:35:23,177] DEEPMD INFO Batch 14600: total wall time = 19.52 s, eta = 0:08:46
[2026-06-20 22:35:43,261] DEEPMD INFO Batch 14800: trn: rmse = 5.01e-01, rmse_e = 4.95e-04, rmse_f = 4.99e-02, lr = 1.00e-04
[2026-06-20 22:35:43,268] DEEPMD INFO Batch 14800: total wall time = 20.09 s, eta = 0:08:42
[2026-06-20 22:36:03,268] DEEPMD INFO Batch 15000: trn: rmse = 4.31e-01, rmse_e = 9.64e-04, rmse_f = 4.29e-02, lr = 1.00e-04
[2026-06-20 22:36:03,271] DEEPMD INFO Batch 15000: total wall time = 20.00 s, eta = 0:08:20
[2026-06-20 22:36:03,320] DEEPMD INFO Saved model to model.ckpt-15000.pt
[2026-06-20 22:36:22,786] DEEPMD INFO Batch 15200: trn: rmse = 4.91e-01, rmse_e = 3.39e-03, rmse_f = 4.88e-02, lr = 1.00e-04
[2026-06-20 22:36:22,793] DEEPMD INFO Batch 15200: total wall time = 19.52 s, eta = 0:07:48
[2026-06-20 22:36:42,512] DEEPMD INFO Batch 15400: trn: rmse = 4.81e-01, rmse_e = 2.04e-04, rmse_f = 4.79e-02, lr = 1.00e-04
[2026-06-20 22:36:42,519] DEEPMD INFO Batch 15400: total wall time = 19.73 s, eta = 0:07:33
[2026-06-20 22:37:02,050] DEEPMD INFO Batch 15600: trn: rmse = 4.09e-01, rmse_e = 2.76e-04, rmse_f = 4.07e-02, lr = 1.00e-04
[2026-06-20 22:37:02,057] DEEPMD INFO Batch 15600: total wall time = 19.54 s, eta = 0:07:09
[2026-06-20 22:37:21,784] DEEPMD INFO Batch 15800: trn: rmse = 4.85e-01, rmse_e = 2.30e-03, rmse_f = 4.82e-02, lr = 1.00e-04
[2026-06-20 22:37:21,790] DEEPMD INFO Batch 15800: total wall time = 19.73 s, eta = 0:06:54
[2026-06-20 22:37:41,545] DEEPMD INFO Batch 16000: trn: rmse = 4.81e-01, rmse_e = 2.93e-03, rmse_f = 4.79e-02, lr = 1.00e-04
[2026-06-20 22:37:41,550] DEEPMD INFO Batch 16000: total wall time = 19.76 s, eta = 0:06:35
[2026-06-20 22:38:01,099] DEEPMD INFO Batch 16200: trn: rmse = 4.87e-01, rmse_e = 3.10e-03, rmse_f = 4.85e-02, lr = 1.00e-04
[2026-06-20 22:38:01,108] DEEPMD INFO Batch 16200: total wall time = 19.56 s, eta = 0:06:11
[2026-06-20 22:38:21,186] DEEPMD INFO Batch 16400: trn: rmse = 4.69e-01, rmse_e = 7.33e-03, rmse_f = 4.66e-02, lr = 1.00e-04
[2026-06-20 22:38:21,190] DEEPMD INFO Batch 16400: total wall time = 20.08 s, eta = 0:06:01
[2026-06-20 22:38:40,470] DEEPMD INFO Batch 16600: trn: rmse = 4.29e-01, rmse_e = 7.30e-03, rmse_f = 4.26e-02, lr = 1.00e-04
[2026-06-20 22:38:40,477] DEEPMD INFO Batch 16600: total wall time = 19.29 s, eta = 0:05:27
[2026-06-20 22:39:00,172] DEEPMD INFO Batch 16800: trn: rmse = 4.33e-01, rmse_e = 4.23e-04, rmse_f = 4.31e-02, lr = 1.00e-04
[2026-06-20 22:39:00,180] DEEPMD INFO Batch 16800: total wall time = 19.70 s, eta = 0:05:15
[2026-06-20 22:39:20,373] DEEPMD INFO Batch 17000: trn: rmse = 4.52e-01, rmse_e = 2.18e-03, rmse_f = 4.50e-02, lr = 1.00e-04
[2026-06-20 22:39:20,381] DEEPMD INFO Batch 17000: total wall time = 20.20 s, eta = 0:05:03
[2026-06-20 22:39:40,308] DEEPMD INFO Batch 17200: trn: rmse = 4.08e-01, rmse_e = 3.13e-03, rmse_f = 4.06e-02, lr = 1.00e-04
[2026-06-20 22:39:40,316] DEEPMD INFO Batch 17200: total wall time = 19.93 s, eta = 0:04:39
[2026-06-20 22:40:00,114] DEEPMD INFO Batch 17400: trn: rmse = 4.41e-01, rmse_e = 2.69e-03, rmse_f = 4.39e-02, lr = 1.00e-04
[2026-06-20 22:40:00,119] DEEPMD INFO Batch 17400: total wall time = 19.80 s, eta = 0:04:17
[2026-06-20 22:40:19,984] DEEPMD INFO Batch 17600: trn: rmse = 4.57e-01, rmse_e = 4.51e-03, rmse_f = 4.55e-02, lr = 1.00e-04
[2026-06-20 22:40:19,993] DEEPMD INFO Batch 17600: total wall time = 19.87 s, eta = 0:03:58
[2026-06-20 22:40:39,371] DEEPMD INFO Batch 17800: trn: rmse = 3.82e-01, rmse_e = 2.84e-03, rmse_f = 3.80e-02, lr = 1.00e-04
[2026-06-20 22:40:39,377] DEEPMD INFO Batch 17800: total wall time = 19.38 s, eta = 0:03:33
[2026-06-20 22:40:58,791] DEEPMD INFO Batch 18000: trn: rmse = 4.05e-01, rmse_e = 7.49e-04, rmse_f = 4.03e-02, lr = 1.00e-04
[2026-06-20 22:40:58,798] DEEPMD INFO Batch 18000: total wall time = 19.42 s, eta = 0:03:14
[2026-06-20 22:41:18,861] DEEPMD INFO Batch 18200: trn: rmse = 4.47e-01, rmse_e = 4.06e-03, rmse_f = 4.45e-02, lr = 1.00e-04
[2026-06-20 22:41:18,867] DEEPMD INFO Batch 18200: total wall time = 20.07 s, eta = 0:03:00
[2026-06-20 22:41:38,563] DEEPMD INFO Batch 18400: trn: rmse = 4.09e-01, rmse_e = 8.48e-04, rmse_f = 4.07e-02, lr = 1.00e-04
[2026-06-20 22:41:38,567] DEEPMD INFO Batch 18400: total wall time = 19.70 s, eta = 0:02:37
[2026-06-20 22:41:58,049] DEEPMD INFO Batch 18600: trn: rmse = 5.17e-01, rmse_e = 6.51e-04, rmse_f = 5.15e-02, lr = 1.00e-04
[2026-06-20 22:41:58,054] DEEPMD INFO Batch 18600: total wall time = 19.49 s, eta = 0:02:16
[2026-06-20 22:42:17,902] DEEPMD INFO Batch 18800: trn: rmse = 4.40e-01, rmse_e = 9.50e-04, rmse_f = 4.38e-02, lr = 1.00e-04
[2026-06-20 22:42:17,908] DEEPMD INFO Batch 18800: total wall time = 19.85 s, eta = 0:01:59
[2026-06-20 22:42:37,099] DEEPMD INFO Batch 19000: trn: rmse = 4.25e-01, rmse_e = 5.43e-03, rmse_f = 4.22e-02, lr = 1.00e-04
[2026-06-20 22:42:37,106] DEEPMD INFO Batch 19000: total wall time = 19.20 s, eta = 0:01:35
[2026-06-20 22:42:56,673] DEEPMD INFO Batch 19200: trn: rmse = 3.98e-01, rmse_e = 3.20e-04, rmse_f = 3.96e-02, lr = 1.00e-04
[2026-06-20 22:42:56,677] DEEPMD INFO Batch 19200: total wall time = 19.57 s, eta = 0:01:18
[2026-06-20 22:43:16,405] DEEPMD INFO Batch 19400: trn: rmse = 3.84e-01, rmse_e = 3.85e-03, rmse_f = 3.82e-02, lr = 1.00e-04
[2026-06-20 22:43:16,412] DEEPMD INFO Batch 19400: total wall time = 19.74 s, eta = 0:00:59
[2026-06-20 22:43:35,832] DEEPMD INFO Batch 19600: trn: rmse = 4.39e-01, rmse_e = 4.60e-03, rmse_f = 4.36e-02, lr = 1.00e-04
[2026-06-20 22:43:35,839] DEEPMD INFO Batch 19600: total wall time = 19.43 s, eta = 0:00:38
[2026-06-20 22:43:55,557] DEEPMD INFO Batch 19800: trn: rmse = 4.25e-01, rmse_e = 1.40e-03, rmse_f = 4.23e-02, lr = 1.00e-04
[2026-06-20 22:43:55,560] DEEPMD INFO Batch 19800: total wall time = 19.72 s, eta = 0:00:19
[2026-06-20 22:44:15,194] DEEPMD INFO Batch 20000: trn: rmse = 4.03e-01, rmse_e = 7.60e-03, rmse_f = 4.00e-02, lr = 1.00e-04
[2026-06-20 22:44:15,201] DEEPMD INFO Batch 20000: total wall time = 19.64 s, eta = 0:00:00
[2026-06-20 22:44:15,260] DEEPMD INFO Saved model to model.ckpt-20000.pt
[2026-06-20 22:44:15,266] DEEPMD INFO average training time: 0.0986 s/batch (200 batches excluded)
[2026-06-20 22:44:15,266] DEEPMD INFO Trained model has been saved to: model.ckpt
3.7. Plot the training curve#
DeePMD-kit writes lcurve.out during training. This cell plots the total RMSE, energy RMSE, and force RMSE when the corresponding columns are present.
import matplotlib.pyplot as plt
from pathlib import Path
import numpy as np
student_workdir = Path("runs/cyclohexane_dpa3_distill/student_se_atten_v2")
lcurve_path = student_workdir / "lcurve.out"
if not lcurve_path.exists():
raise FileNotFoundError(f"Training curve file not found: {lcurve_path}")
with lcurve_path.open("r", encoding="utf-8") as f:
header = f.readline().lstrip("#").split()
lcurve = np.atleast_2d(np.loadtxt(lcurve_path))
lcurve_data = {name: lcurve[:, idx] for idx, name in enumerate(header)}
steps = lcurve_data["step"]
metrics = [
("rmse_trn", "Total RMSE", "RMSE"),
("rmse_e_trn", "Energy RMSE", "eV"),
("rmse_f_trn", "Force RMSE", "eV/Angstrom"),
]
metrics = [(key, title, ylabel) for key, title, ylabel in metrics if key in lcurve_data]
fig, axes = plt.subplots(1, len(metrics), figsize=(4 * len(metrics), 3.5))
axes = np.atleast_1d(axes)
for ax, (key, title, ylabel) in zip(axes, metrics, strict=False):
ax.plot(steps, lcurve_data[key], marker="o", label="train")
ax.set_title(title)
ax.set_xlabel("step")
ax.set_xscale("log")
ax.set_ylabel(ylabel)
ax.set_yscale("log")
ax.grid(True, alpha=0.3)
ax.legend()
fig.tight_layout()
plt.show()
print("lcurve:", lcurve_path)
print("columns:", header)
print("steps:", int(steps[0]), "->", int(steps[-1]))
lcurve: runs/cyclohexane_dpa3_distill/student_se_atten_v2/lcurve.out
columns: ['step', 'rmse_trn', 'rmse_e_trn', 'rmse_f_trn', 'lr']
steps: 1 -> 20000
3.8. Freeze and evaluate the Student model on the distillation dataset#
After training, export the model and evaluate it against the Teacher labels in the distillation dataset.
Step 1: use dp --pt freeze to convert the trained checkpoint into a frozen model file under artifacts/models/.
Step 2: use dp --pt test on data/cyclohexane/train/system_001 to report energy and force errors. This is a training-set distillation error, not an independent test-set accuracy.
Run dp --pt freeze to obtain student_se_atten_v2.pth, then compress the exported model.
%%bash
set -euo pipefail
STUDENT_WORKDIR="runs/cyclohexane_dpa3_distill/student_se_atten_v2"
STUDENT_MODEL="$(pwd)/runs/cyclohexane_dpa3_distill/artifacts/models/student_se_atten_v2.pth"
COMPRESSED_MODEL="$(pwd)/runs/cyclohexane_dpa3_distill/artifacts/models/student_compressed.pth"
FREEZE_LOG="runs/cyclohexane_dpa3_distill/artifacts/reports/freeze.log"
mkdir -p "$(dirname "$STUDENT_MODEL")" "$(dirname "$FREEZE_LOG")"
(
cd "$STUDENT_WORKDIR"
dp --pt freeze -o "$STUDENT_MODEL"
) > "$FREEZE_LOG" 2>&1
echo "freeze finished: $STUDENT_MODEL"
echo "freeze log: $FREEZE_LOG"
dp --pt compress -i "$STUDENT_MODEL" -o "$COMPRESSED_MODEL"
freeze finished: //runs/cyclohexane_dpa3_distill/artifacts/models/student_se_atten_v2.pth
freeze log: runs/cyclohexane_dpa3_distill/artifacts/reports/freeze.log
To get the best performance, it is recommended to adjust the number of threads by setting the environment variables OMP_NUM_THREADS, DP_INTRA_OP_PARALLELISM_THREADS, and DP_INTER_OP_PARALLELISM_THREADS. See https://deepmd.rtfd.io/parallelism/ for more information.
[2026-06-20 22:44:44,455] DEEPMD INFO DeePMD version: 3.1.3
[2026-06-20 22:44:44,879] DEEPMD INFO training data with lower boundary: [[-1.1939386 -0. -0. -0. ]
[-0.9941983 -0. -0. -0. ]]
[2026-06-20 22:44:44,880] DEEPMD INFO training data with upper boundary: [[2.3226078 3.9109068 3.9109068 3.9109068]
[3.28381 5.254691 5.254691 5.254691 ]]
Run dp --pt test on the distillation dataset.
%%bash
set -euo pipefail
COMPRESSED_MODEL="$(pwd)/runs/cyclohexane_dpa3_distill/artifacts/models/student_compressed.pth"
TRAIN_SYSTEM="$(pwd)/data/cyclohexane/train/system_001"
TEST_LOG="runs/cyclohexane_dpa3_distill/artifacts/reports/student_train_eval.log"
N_EVAL_FRAMES=1000
mkdir -p "$(dirname "$TEST_LOG")"
dp --pt test -m "$COMPRESSED_MODEL" -s "$TRAIN_SYSTEM" -n "$N_EVAL_FRAMES" > "$TEST_LOG" 2>&1
echo "training-set evaluation finished; log: $TEST_LOG"
echo
echo "Training-set distillation error:"
grep -E "number of test data|Energy MAE|Energy RMSE|Force MAE|Force RMSE" "$TEST_LOG" | awk '!seen[$0]++'
training-set evaluation finished; log: runs/cyclohexane_dpa3_distill/artifacts/reports/student_train_eval.log
Training-set distillation error:
[2026-06-20 22:45:02,675] DEEPMD INFO # number of test data : 501
[2026-06-20 22:45:02,677] DEEPMD INFO Energy MAE : 1.300363e-01 eV
[2026-06-20 22:45:02,677] DEEPMD INFO Energy RMSE : 1.301760e-01 eV
[2026-06-20 22:45:02,677] DEEPMD INFO Energy MAE/Natoms : 7.224239e-03 eV
[2026-06-20 22:45:02,677] DEEPMD INFO Energy RMSE/Natoms : 7.231999e-03 eV
[2026-06-20 22:45:02,677] DEEPMD INFO Force MAE : 3.166588e-02 eV/Å
[2026-06-20 22:45:02,677] DEEPMD INFO Force RMSE : 4.142549e-02 eV/Å
[2026-06-20 22:45:02,678] DEEPMD INFO Energy MAE : 1.300363e-01 eV
[2026-06-20 22:45:02,678] DEEPMD INFO Energy RMSE : 1.301760e-01 eV
[2026-06-20 22:45:02,678] DEEPMD INFO Energy MAE/Natoms : 7.224239e-03 eV
[2026-06-20 22:45:02,678] DEEPMD INFO Energy RMSE/Natoms : 7.231999e-03 eV
[2026-06-20 22:45:02,678] DEEPMD INFO Force MAE : 3.166588e-02 eV/Å
[2026-06-20 22:45:02,678] DEEPMD INFO Force RMSE : 4.142549e-02 eV/Å
3.9. Run MD with the distilled Student model#
The frozen and compressed Student model can now be used as an ASE calculator. The default Student MD length is 40,000 steps with a 0.5 fs time step, saving one frame every 100 steps. This gives a longer post-training trajectory while keeping the notebook output size manageable.
STUDENT_MODEL = Path(
"runs/cyclohexane_dpa3_distill/artifacts/models/student_compressed.pth"
)
STUDENT_MD_TEMPERATURE_K = 300
STUDENT_MD_TIMESTEP_FS = 0.5
STUDENT_MD_STEPS = 40000
STUDENT_SAVE_EVERY = 100
initial_xyz = Path("data/cyclohexane/raw/cyclohexane_initial.xyz")
student_md_xyz = Path("data/cyclohexane/raw/cyclohexane_student.xyz")
from ase.io import write, read
from tqdm import trange
from deepmd.calculator import DP
from ase import units
from ase.md.velocitydistribution import MaxwellBoltzmannDistribution
from ase.md.verlet import VelocityVerlet
student_atoms = read(str(initial_xyz))
student_atoms.calc = DP(model=str(STUDENT_MODEL))
rng = np.random.default_rng(20260507)
MaxwellBoltzmannDistribution(
student_atoms, temperature_K=STUDENT_MD_TEMPERATURE_K, rng=rng
)
student_dynamics = VelocityVerlet(
student_atoms, timestep=STUDENT_MD_TIMESTEP_FS * units.fs
)
student_md_frames = [student_atoms.copy()]
student_md_energies = [float(student_atoms.get_potential_energy())]
print(
f"Running Student MD for {STUDENT_MD_STEPS} steps, "
f"timestep={STUDENT_MD_TIMESTEP_FS} fs, T={STUDENT_MD_TEMPERATURE_K} K"
)
for step in trange(1, STUDENT_MD_STEPS + 1):
student_dynamics.run(1)
if step % STUDENT_SAVE_EVERY == 0:
student_md_frames.append(student_atoms.copy())
student_md_energies.append(float(student_atoms.get_potential_energy()))
write(str(student_md_xyz), student_md_frames)
print("Student MD finished")
print("saved frames:", len(student_md_frames))
print("trajectory:", student_md_xyz)
print("energy range:", min(student_md_energies), max(student_md_energies), "eV")
You can use the environment variable DP_INFER_BATCH_SIZE tocontrol the inference batch size (nframes * natoms). The default value is 1024.
Running Student MD for 40000 steps, timestep=0.5 fs, T=300 K
100%|██████████| 40000/40000 [06:19<00:00, 105.40it/s]
Student MD finished
saved frames: 401
trajectory: data/cyclohexane/raw/cyclohexane_student.xyz
energy range: -6417.227179962489 -6416.89055849919 eV
3.10. Visualize representative Student MD structures#
This section uses the student_md_frames and student_md_energies generated above. It selects the lowest-energy and highest-energy saved frames as simple representatives from the Student MD trajectory.
The energy values are the Student model potential energies for the full cyclohexane system in eV. They are not DFT energies and should not be interpreted as a rigorous conformer ranking.
For cyclohexane, a stable room-temperature trajectory is expected to show chair-like structures most of the time. The highest-energy frame in this short trajectory is better described as a distorted structure within the sampled region, not necessarily a true transition-state or boat conformer. To study chair-boat-ring-inversion statistics, run longer sampling, use higher-temperature or enhanced-sampling MD, and validate the resulting structures separately.
import matplotlib.pyplot as plt
plt.plot(range(len(student_md_energies)), student_md_energies)
plt.xlabel("saved frame index")
plt.ylabel("Student potential energy (eV)")
plt.title("Student MD energy trace")
plt.grid(True, alpha=0.3)
plt.show()
student_md_energies = np.asarray(student_md_energies)
low_idx = int(np.argmin(student_md_energies))
high_idx = int(np.argmax(student_md_energies))
print("trajectory frames:", len(student_md_frames))
print("lowest-energy frame:", low_idx, f"{student_md_energies[low_idx]:.6f} eV")
print("highest-energy frame:", high_idx, f"{student_md_energies[high_idx]:.6f} eV")
view_style = {
"stick": {"radius": 0.12},
"sphere": {"radius": 0.35},
}
for label, frame_idx in [
("lowest energy", low_idx),
("highest energy", high_idx),
]:
print(f"{label}: frame {frame_idx}, E = {student_md_energies[frame_idx]:.6f} eV")
view = dpdata.System(student_md_frames[frame_idx], fmt="ase/structure").to_3dmol(
f_idx=0,
size=(450, 350),
style=view_style,
)
view.show()
trajectory frames: 401
lowest-energy frame: 0 -6417.227180 eV
highest-energy frame: 137 -6416.890558 eV
lowest energy: frame 0, E = -6417.227180 eV
3Dmol.js failed to load for some reason. Please check your browser console for error messages.
highest energy: frame 137, E = -6416.890558 eV
3Dmol.js failed to load for some reason. Please check your browser console for error messages.