yli697 commited on
Commit
eacf032
·
verified ·
1 Parent(s): 45eacc9

Release LUMINA-2M v0.1.4 (2.4M params)

Browse files
Files changed (4) hide show
  1. README.md +9 -5
  2. model.pt +2 -2
  3. model.safetensors +1 -1
  4. pareto_front.png +0 -0
README.md CHANGED
@@ -21,16 +21,16 @@ Trained on ACOPF datasets spanning **multiple network topologies**, LUMINA-2M ge
21
 
22
  ## Model Details
23
 
24
- - **Version**: v0.1.3
25
  - **Model Architecture**: HGT
26
  - **Total Parameters**: 2,375,548 (2.4M)
27
  - **Trainable Parameters**: 2,375,548 (2.4M)
28
  - **Training Case**: case14_ieee, case30_ieee, case57_ieee, case118_ieee, case500_goc, case2000_goc, case4661_sdet, case6470_rte, case10000_goc, case13659_pegase
29
  - **Training Data Size**: 240,000 samples per case
30
- - **Training Date**: 2026-08-12
31
- - **Training Epochs**: 69
32
  - **Loss Type**: Augmented Lagrangian
33
- - **Final Validation Loss**: 0.190340
34
 
35
  > **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.
36
 
@@ -47,6 +47,7 @@ This repository contains the model in multiple formats:
47
  - `model.pt` - Complete PyTorch checkpoint with metadata and config
48
  - `model.safetensors` - Model weights in SafeTensors format (recommended)
49
  - `requirements.txt` - Required Python libraries
 
50
 
51
  ## Installation
52
 
@@ -158,8 +159,11 @@ https://doi.org/10.48550/arXiv.2406.07234.
158
 
159
  ## Release History
160
 
 
 
161
  | Version | Date | Epochs | Val Loss | Training Case |
162
  |---------|------|--------|----------|---------------|
 
163
  | [v0.1.3](https://huggingface.co/argonne/LUMINA-2M/tree/v0.1.3) | 2026-08-12 | 69 | 0.190340 | full-topology + N-1 + FT |
164
  | [v0.1.2](https://huggingface.co/argonne/LUMINA-2M/tree/v0.1.2) | 2026-07-31 | 63 | 0.120890 | full-topology + N-1 |
165
  | [v0.1.1](https://huggingface.co/argonne/LUMINA-2M/tree/v0.1.1) | 2026-07-16 | 34 | 0.052400 | case14_ieee, case30_ieee, case57_ieee, case118_ieee, case500_goc, case2000_goc, case4661_sdet, case6470_rte, case10000_goc, case13659_pegase |
@@ -193,6 +197,6 @@ This research used resources of the National Energy Research Scientific Computin
193
 
194
  ## License
195
 
196
- Released under the Apache License, Version 2.0. See the bundled [`LICENSE`](https://huggingface.co/argonne/LUMINA-2M/blob/v0.1.3/LICENSE) and [`NOTICE`](https://huggingface.co/argonne/LUMINA-2M/blob/v0.1.3/NOTICE) files for the full terms and attribution.
197
 
198
  Copyright 2026 UChicago Argonne, LLC.
 
21
 
22
  ## Model Details
23
 
24
+ - **Version**: v0.1.4
25
  - **Model Architecture**: HGT
26
  - **Total Parameters**: 2,375,548 (2.4M)
27
  - **Trainable Parameters**: 2,375,548 (2.4M)
28
  - **Training Case**: case14_ieee, case30_ieee, case57_ieee, case118_ieee, case500_goc, case2000_goc, case4661_sdet, case6470_rte, case10000_goc, case13659_pegase
29
  - **Training Data Size**: 240,000 samples per case
30
+ - **Training Date**: 2026-08-27
31
+ - **Training Epochs**: 71
32
  - **Loss Type**: Augmented Lagrangian
33
+ - **Final Validation Loss**: 0.531175
34
 
35
  > **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.
36
 
 
47
  - `model.pt` - Complete PyTorch checkpoint with metadata and config
48
  - `model.safetensors` - Model weights in SafeTensors format (recommended)
49
  - `requirements.txt` - Required Python libraries
50
+ - `pareto_front.png` - Accuracy vs. constraint-violation plot across all released checkpoints (see Release History)
51
 
52
  ## Installation
53
 
 
159
 
160
  ## Release History
161
 
162
+ ![Accuracy vs. constraint violation across released checkpoints](pareto_front.png)
163
+
164
  | Version | Date | Epochs | Val Loss | Training Case |
165
  |---------|------|--------|----------|---------------|
166
+ | [v0.1.4](https://huggingface.co/argonne/LUMINA-2M/tree/v0.1.4) | 2026-08-27 | 71 | 0.531175 | full-topology + N-1 + FT |
167
  | [v0.1.3](https://huggingface.co/argonne/LUMINA-2M/tree/v0.1.3) | 2026-08-12 | 69 | 0.190340 | full-topology + N-1 + FT |
168
  | [v0.1.2](https://huggingface.co/argonne/LUMINA-2M/tree/v0.1.2) | 2026-07-31 | 63 | 0.120890 | full-topology + N-1 |
169
  | [v0.1.1](https://huggingface.co/argonne/LUMINA-2M/tree/v0.1.1) | 2026-07-16 | 34 | 0.052400 | case14_ieee, case30_ieee, case57_ieee, case118_ieee, case500_goc, case2000_goc, case4661_sdet, case6470_rte, case10000_goc, case13659_pegase |
 
197
 
198
  ## License
199
 
200
+ Released under the Apache License, Version 2.0. See the bundled [`LICENSE`](https://huggingface.co/argonne/LUMINA-2M/blob/v0.1.4/LICENSE) and [`NOTICE`](https://huggingface.co/argonne/LUMINA-2M/blob/v0.1.4/NOTICE) files for the full terms and attribution.
201
 
202
  Copyright 2026 UChicago Argonne, LLC.
model.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:36de7fd5d1de3aee8020999b35ce0bb958914159c2a198ccf0ed727bd1acd8cb
3
- size 28499255
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0a01fc984821090cefe1de24218790caa0dbcc70d6ff71b670eae19716906669
3
+ size 28499319
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:060ae5f7b6fd525a1495287dd6c67e661cb04f8b7d84370a2cdfc68ddab529ca
3
  size 9520848
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:002e3faea331731d2fcd5bae16db8508bf91ba2125532af0dee41f0dbcace03d
3
  size 9520848
pareto_front.png ADDED