ConcPre / README.md
Heterogeneity2025's picture
Upload 32 files
7a1376e verified
|
Raw
History Blame Contribute Delete
1.06 kB
metadata
title: Concrete Strength Predict Design
colorFrom: gray
colorTo: blue
sdk: docker
app_port: 8501
pinned: false
license: mit

Concrete Compressive-Strength Predict / Design

An interactive tool for concrete mix design.

  • Predict strength - enter a mix design, with optional fields allowed, and get the predicted compressive strength.
  • Design a mix - enter a target strength and get several detailed mix designs that reach it.

Trained on about 9,700 mixes spanning about 2 to 198 MPa, from normal-strength concrete through ultra-high-performance concrete.

Running locally

pip install -r requirements.txt
streamlit run streamlit_app.py

Files

file purpose
streamlit_app.py the web UI
inference.py model loading plus predict / design logic
build_inverse_index.py offline script to precompute the design index and model_config.json
checkpoints_full_rich/ trained model files and model_config.json
inverse_index.csv precomputed predictions over the training mixes