Kiarash99/GPTMicro-Nanowire-Sintering

GPTMicro — Nanowire Sintering & Symbolic Regression Dataset

Curated data for data-driven discovery of governing equations in nanowire sintering. It pairs raw molecular-dynamics (MD) trajectories with the ML-ready train/validation/test splits used to learn closed-form models for the sintering dynamics (change in flattening ddelta and rotation dtheta) and for two effective material properties (effective diffusion coefficient D_eff and effective relaxation/viscosity coefficient tau_eff).

Code

The full GPT-Micro implementation that produces and consumes this data lives here:

💻 https://github.com/kiarash99Naghavi/GPT-Micro

GPT-Micro is a large-language-model paradigm for accelerated, inexpensive, and thermodynamics-consistent discovery of constitutive models in manufacturing. It runs a two-stage autonomous loop: Stage 1 uses a Retrieval-Augmented-Generation (RAG) agent to mine legacy literature and an LLM to propose, calibrate, and iteratively refine analytical hypotheses for the state–microstructure models (ddelta, dtheta); Stage 2 combines those with thermodynamic conservation laws to synthesize (property, state) data and runs PySR symbolic regression to discover compact closed-form constitutive models for the effective properties (D_eff, tau_eff). The repository README explains how to drop the folders from this dataset into its data/ directory to reproduce the four nanowire-sintering case studies.

Parameter grid

Each MD trajectory is a sweep over three design parameters:

That is 5 × 5 × 5 = 125 trajectories, each sampled at 21 time points (t = 0 → 1000 ps, step 50 ps).

Contents

GPTMicro-HF-Dataset/
├── raw_md_trajectories/            # 125 raw MD trajectory CSVs (shared source)
│   └── R_<radius>_T_<temp>_theta0_<angle>.csv
├── nanowire_ddelta/                # ML splits, target = ddelta
│   ├── training.csv                # 567 rows
│   ├── validation.csv              # 567 rows
│   ├── testing.csv                 # 1,491 rows
│   └── testing_extrapolation.csv   # 1,575 rows (out-of-range R/T)
├── nanowire_dtheta/                # ML splits, target = dtheta
│   ├── training.csv                # 567 rows
│   ├── validation.csv              # 567 rows
│   ├── testing.csv                 # 1,491 rows
│   └── testing_extrapolation.csv   # 1,575 rows
├── D_eff_symbolic_regression/      # effective diffusion coefficient
│   ├── train_data.xlsx / validation_data.xlsx / test_data.xlsx
│   ├── combined_D_eff_results.xlsx # full R/T/theta0/time table
│   └── R_with_x1_x2_x3.xlsx        # per-radius engineered features
└── tau_eff_symbolic_regression/    # effective relaxation/viscosity coefficient
    ├── train_data.xlsx / validation_data.xlsx / test_data.xlsx
    ├── combined_tau_eff_results.xlsx
    └── R_with_x1_x2_x3.xlsx

Column reference

*`raw_md_trajectories/.csv`** — one sintering trajectory per file:

*`nanowire_ddelta/.csv** (target ddelta`):

*`nanowire_dtheta/.csv** (target dtheta): identical schema with theta0deg`, `theta1deg, and target dtheta`.

*`symbolicregression/traindata.xlsx`** (Deff example): R, T, theta_0, t, D_eff_nm2_s_raw, D_eff (scaled target). The tau_eff tables use tau_eff_kg/s as the target. R_with_x1_x2_x3.xlsx provides engineered per-radius features x1, x2, x3.

Splits

• —train / validation / test — random split over the in-range parameter grid.
• —test_extrapolation — held-out trajectories at parameter values outside the training range (e.g. R = 5), for evaluating out-of-distribution generalization of discovered equations.

Usage

Load a nanowire task directly with the 🤗 datasets library:

from datasets import load_dataset

ds = load_dataset("Kiarash99/GPTMicro-Nanowire-Sintering", "nanowire_ddelta")
print(ds)
# DatasetDict({ train, validation, test, test_extrapolation })

row = ds["train"][0]
print(row)   # {'R': 7.5, 'T': 400, 'theta_0': 0, 't': 0, 'delta0': 0.1, ...}

Read a raw trajectory or the symbolic-regression tables with pandas:

import pandas as pd
traj = pd.read_csv("raw_md_trajectories/R_10_T_600_theta0_30.csv")
deff = pd.read_excel("D_eff_symbolic_regression/train_data.xlsx")

License

Released under the MIT License (see LICENSE).

Citation

If you use this dataset, please cite the associated work:

@misc{gptmicro_nanowire_sintering,
  title  = {GPTMicro: Nanowire Sintering & Symbolic Regression Dataset},
  author = {Dutta, Soumik and Khanghah, Kiarash Naghavi and Shree, Sania and McNeil, Logan and Feldhausen, Thomas and Xu, Hongyi and Malhotra, Rajiv},
  year   = {2026},
  note   = {Hugging Face Datasets}
}