Release LUMINA-2M v0.1.5 (3.8M params)

README.md

@@ -14,23 +14,23 @@ license: apache-2.0
 
 ## Model Description
 
-**LUMINA-2M** is a heterogeneous graph neural network (HGT) designed for **fast, high-fidelity prediction of AC Optimal Power Flow (ACOPF) solutions**.
+**LUMINA-2M** is a 3.8M-parameter heterogeneous graph neural network (HGT) designed for **fast, high-fidelity prediction of AC Optimal Power Flow (ACOPF) solutions**.
 The model encodes power system components—buses, generators, loads, and transmission lines—as distinct node and edge types, capturing structural and physical heterogeneity in electric grids.
 
 Trained on ACOPF datasets spanning **multiple network topologies**, LUMINA-2M generalizes across diverse grid configurations and operating conditions. It provides efficient approximations of OPF variables such as voltages, power injections, and line flows, serving as a scalable surrogate for traditional optimization solvers.
 
 ## Model Details
 
-- **Version**: v0.1.5-alpha
+- **Version**: v0.1.5
 - **Model Architecture**: HGT
 - **Total Parameters**: 3,755,642 (3.8M)
 - **Trainable Parameters**: 3,755,642 (3.8M)
 - **Training Case**: case14_ieee, case30_ieee, case57_ieee, case118_ieee, case500_goc, case2000_goc, case4661_sdet, case6470_rte, case10000_goc, case13659_pegase
-- **Training Data Size**: 270,000 samples per case
-- **Training Date**: 2026-09-23
-- **Training Epochs**: 86
+- **Training Data Size**: 240,000 samples per case
+- **Training Date**: 2026-10-09
+- **Training Epochs**: 87
 - **Loss Type**: Augmented Lagrangian
-- **Final Validation Loss**: 0.072916
+- **Final Validation Loss**: 0.664171
 
 > **Note**: The validation loss is only comparable across releases that share the same **Loss Type**. For example, an Augmented Lagrangian loss includes constraint-violation penalties and Lagrangian terms, so its magnitude is not directly comparable to a plain MSE loss.
 
@@ -164,6 +164,7 @@ https://doi.org/10.48550/arXiv.2406.07234.
 
 | Version | Date | Epochs | Val Loss | Training Case |
 |---------|------|--------|----------|---------------|
+| [v0.1.5](https://huggingface.co/argonne/LUMINA-2M/tree/v0.1.5) | 2026-10-09 | 87 | 0.664171 | case14_ieee, case30_ieee, case57_ieee, case118_ieee, case500_goc, case2000_goc, case4661_sdet, case6470_rte, case10000_goc, case13659_pegase |
 | [v0.1.5-alpha](https://huggingface.co/argonne/LUMINA-2M/tree/v0.1.5-alpha) | 2026-09-23 | 86 | 0.072916 | full-topology |
 | [v0.1.4](https://huggingface.co/argonne/LUMINA-2M/tree/v0.1.4) | 2026-08-27 | 71 | 0.531175 | full-topology + N-1 + FT |
 | [v0.1.3](https://huggingface.co/argonne/LUMINA-2M/tree/v0.1.3) | 2026-08-12 | 69 | 0.190340 | full-topology + N-1 + FT |
@@ -199,6 +200,6 @@ This research used resources of the National Energy Research Scientific Computin
 
 ## License
 
-Released under the Apache License, Version 2.0. See the bundled [`LICENSE`](https://huggingface.co/argonne/LUMINA-2M/blob/v0.1.5-alpha/LICENSE) and [`NOTICE`](https://huggingface.co/argonne/LUMINA-2M/blob/v0.1.5-alpha/NOTICE) files for the full terms and attribution.
+Released under the Apache License, Version 2.0. See the bundled [`LICENSE`](https://huggingface.co/argonne/LUMINA-2M/blob/v0.1.5/LICENSE) and [`NOTICE`](https://huggingface.co/argonne/LUMINA-2M/blob/v0.1.5/NOTICE) files for the full terms and attribution.
 
 Copyright 2026 UChicago Argonne, LLC.

config.json

@@ -110,12 +110,13 @@
   "config": {
     "models": {
       "HGT": {
-        "hidden_channels": 128,
-        "num_layers": 6,
-        "num_heads": 2,
         "backend": "sage",
-        "dropout": 0.0
+        "num_layers": 6,
+        "hidden_channels": 128,
+        "dropout": 0.24398355775520905,
+        "num_heads": 2
       }
     }
-  }
+  },
+  "minmax_scaling": true
 }

model.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:70bf1593a0200526cda7d0838358044c86e08685780631be6a05d781ea9e3fdf
-size 45051079
+oid sha256:47b7862e2629c65542feea6dfb8ada45a6d7527b5637a2604388be4bee499e20
+size 45049132

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6872d607090a5b724d16b7799667041cb35d1a43a99808075b70eb32213911d2
+oid sha256:543f2b3b18f88019bff4f971dd1998af251485be0d00f22c9fe2f2d7766da6e6
 size 15050568