Add files using upload-large-folder tool
Browse files- .gitignore +4 -0
- README.md +62 -3
- checkpoints/pxrd_cnn_control_checkpoint.pt +3 -0
- checkpoints/xrd_model_82ept35h_best.pth +3 -0
- checkpoints/xrd_model_9rwv1qly_best.pth +3 -0
- checkpoints/xrd_model_cscjfdwk_best.pth +3 -0
- checkpoints/xrd_model_wwuvp1kj_best.pth +3 -0
- metadata/CHECKPOINT_PLAN.md +12 -0
- metadata/STAGED_INVENTORY.json +10 -0
.gitignore
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Local scratch files
|
| 2 |
+
.DS_Store
|
| 3 |
+
*.tmp
|
| 4 |
+
*.log
|
README.md
CHANGED
|
@@ -1,3 +1,62 @@
|
|
| 1 |
-
---
|
| 2 |
-
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: en
|
| 3 |
+
license: mit
|
| 4 |
+
library_name: pytorch
|
| 5 |
+
tags:
|
| 6 |
+
- materials-science
|
| 7 |
+
- x-ray-diffraction
|
| 8 |
+
- crystallography
|
| 9 |
+
- evaluation
|
| 10 |
+
- neurips
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# InformedXRD Case-Study Checkpoints
|
| 14 |
+
|
| 15 |
+
This staging directory is the draft Hugging Face model card for the NeurIPS Evaluations & Datasets artifact release associated with *InformedXRD*.
|
| 16 |
+
|
| 17 |
+
## Contents
|
| 18 |
+
|
| 19 |
+
The intended public model repo contains only the checkpoints needed to reproduce the paper's evaluation case studies:
|
| 20 |
+
|
| 21 |
+
- four ViT checkpoints used in the main ranking and calibration comparisons
|
| 22 |
+
- one legacy replay-era ViT baseline
|
| 23 |
+
- one 1D residual CNN control checkpoint
|
| 24 |
+
|
| 25 |
+
It does **not** contain:
|
| 26 |
+
|
| 27 |
+
- the full companion-paper checkpoint archive
|
| 28 |
+
- training code snapshots
|
| 29 |
+
- benchmark HDF5 payloads
|
| 30 |
+
|
| 31 |
+
## Recommended filenames
|
| 32 |
+
|
| 33 |
+
The staged CNN control checkpoint has been renamed to a neutral, paper-facing artifact name:
|
| 34 |
+
|
| 35 |
+
- `pxrd_cnn_control_checkpoint.pt`
|
| 36 |
+
|
| 37 |
+
The remaining ViT filenames are kept as-is for now because they align with the existing public checkpoint manifest and reviewer notebooks.
|
| 38 |
+
|
| 39 |
+
## Intended use
|
| 40 |
+
|
| 41 |
+
These checkpoints are intended for:
|
| 42 |
+
|
| 43 |
+
- re-running the NeurIPS evaluation protocol on `RRUFF-473` and `RRUFF-325`
|
| 44 |
+
- validating the hierarchy-aware error metric
|
| 45 |
+
- validating the prior-baseline and stratified-analysis claims
|
| 46 |
+
|
| 47 |
+
They are not intended as generic pretrained models for production PXRD workflows.
|
| 48 |
+
|
| 49 |
+
## Provenance
|
| 50 |
+
|
| 51 |
+
The CNN control belongs to the validated in-house Vista lineage documented in:
|
| 52 |
+
|
| 53 |
+
- `docs/review_notes/cnn_rruff325_identification_2026-04-29.md` in the private repo
|
| 54 |
+
|
| 55 |
+
The ViT checkpoints correspond to the released case-study models enumerated in the public checkpoint manifest.
|
| 56 |
+
|
| 57 |
+
## Publication plan
|
| 58 |
+
|
| 59 |
+
1. Create a private Hugging Face model repo.
|
| 60 |
+
2. Upload the staged checkpoints and metadata.
|
| 61 |
+
3. Verify load paths against the public reproducibility manifests.
|
| 62 |
+
4. Switch visibility when the review/release posture is decided.
|
checkpoints/pxrd_cnn_control_checkpoint.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:10898c96a32996180621bba9916e9a25abf9d47e60e03a576b79c9c245e1a743
|
| 3 |
+
size 7863166
|
checkpoints/xrd_model_82ept35h_best.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:83adf71c980ff1e6fa5a656ce0d477d6b8386425ec00acc914e4b1b8c3741027
|
| 3 |
+
size 10642238
|
checkpoints/xrd_model_9rwv1qly_best.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a79a286ed458da2d2420df0c3e6fcba4ea062abd03bb74c03beed37c5220fc8a
|
| 3 |
+
size 10641854
|
checkpoints/xrd_model_cscjfdwk_best.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ef28488ae21b2331b2a4ecfd0036c369d9f3f8bdcdd48e1ad40ce19a386da9e9
|
| 3 |
+
size 10641982
|
checkpoints/xrd_model_wwuvp1kj_best.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0d891d056b21df6517837be1a66bdafa201edc8c94074b65b635e84128cbd1ab
|
| 3 |
+
size 10641854
|
metadata/CHECKPOINT_PLAN.md
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
## Model Artifact Plan
|
| 2 |
+
|
| 3 |
+
Candidate contents:
|
| 4 |
+
- four ViT checkpoints used as NeurIPS case studies
|
| 5 |
+
- one 1D residual CNN control checkpoint (`pxrd_cnn_control_checkpoint.pt`)
|
| 6 |
+
- per-checkpoint role notes
|
| 7 |
+
- model card text
|
| 8 |
+
|
| 9 |
+
Candidate exclusions:
|
| 10 |
+
- unused PRX checkpoints
|
| 11 |
+
- optimizer states or broad training-history dumps unless needed
|
| 12 |
+
- unrelated ablation checkpoints not discussed in the NeurIPS paper
|
metadata/STAGED_INVENTORY.json
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"checkpoints": [
|
| 3 |
+
"pxrd_cnn_control_checkpoint.pt",
|
| 4 |
+
"xrd_model_82ept35h_best.pth",
|
| 5 |
+
"xrd_model_9rwv1qly_best.pth",
|
| 6 |
+
"xrd_model_cscjfdwk_best.pth",
|
| 7 |
+
"xrd_model_wwuvp1kj_best.pth"
|
| 8 |
+
],
|
| 9 |
+
"missing_expected_candidates": []
|
| 10 |
+
}
|