GlobalWellFM — Global Well Log Foundation Model

Author: Abdulrahman Al-Fakih · KFUPM / Geoscience AI Lab Test F1 (weighted): 0.9494 Training wells: Netherlands (NLOG) + Norway (synthetic) + USA (SEG-2016)

First open-source, multi-basin, physics-constrained lithofacies foundation model trained on well logs from 3 countries.

Usage

from huggingface_hub import hf_hub_download
import joblib
model   = joblib.load(hf_hub_download("Aljaser2030/GlobalWellFM", "model/xgb_model.joblib"))
imputer = joblib.load(hf_hub_download("Aljaser2030/GlobalWellFM", "model/imputer.joblib"))
le      = joblib.load(hf_hub_download("Aljaser2030/GlobalWellFM", "model/label_encoder.joblib"))

Citation

Al-Fakih, A. (2026). GlobalWellFM: Physics-Constrained Foundation Model for Global Well Log Lithofacies Classification. Computers & Geosciences.