Create README.md
#2
by
heloise-chomet
- opened
README.md
ADDED
|
@@ -0,0 +1,19 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
# For reference on dataset card metadata, see the spec: https://github.com/huggingface/hub-docs/blob/main/datasetcard.md?plain=1
|
| 3 |
+
# Doc / guide: https://huggingface.co/docs/hub/datasets-cards
|
| 4 |
+
{}
|
| 5 |
+
---
|
| 6 |
+
# SPICE2_curated
|
| 7 |
+
The provided dataset is based on the version 2 SPICE dataset. The SPICE dataset was chosen for its diversity, both in chemical and conformational space,
|
| 8 |
+
compromising of approximately 2 million structures computed at the ωB97M-D3(BJ)/def2-TZVPPD level of theory.
|
| 9 |
+
## Dataset Structure
|
| 10 |
+
The provided dataset is split into a training set and a validation set with a 95/5 split. The split was done per molecular
|
| 11 |
+
SMILES, insuring that different conformations of the same molecule do not appear in both the training and
|
| 12 |
+
validation set. The training set contains 1 737 896 structures and the validation set 87 922 structures.
|
| 13 |
+
## Filtering Process
|
| 14 |
+
The version 2 SPICE dataset was filtered by removing the following structures:
|
| 15 |
+
- **Unphysical structures:** Only keeping structures where all hydrogen atoms have exactly one bond (removing 42 689 structures).
|
| 16 |
+
- **Charged systems:** Removing all charged systems (removing 142 647 structures).
|
| 17 |
+
- **High forces:** Applying a total force filter of 0.1 eV/A and a maximum force filter of 15 eV/A (removing 1 024 structures).
|
| 18 |
+
## Version 2 SPICE dataset source
|
| 19 |
+
https://pubs.acs.org/doi/10.1021/acs.jctc.4c00794
|