Xin Chen commited on
Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
tags:
|
| 4 |
+
- molecule-generation
|
| 5 |
+
- chemistry
|
| 6 |
+
- smiles
|
| 7 |
+
- drug-discovery
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
# RiemannGen checkpoints
|
| 11 |
+
|
| 12 |
+
Trained checkpoints for the [RiemannGen](https://github.com/) project: two
|
| 13 |
+
from-scratch MolMIM-style latent sequence models.
|
| 14 |
+
|
| 15 |
+
- `atom/final-atom.pt` + `atom/vocab_new.txt` -- plain-SMILES model, trained
|
| 16 |
+
on a combined ZINC15 + MOSES + ChEMBL corpus.
|
| 17 |
+
- `fragment/final.pt` + `fragment/vocab_new.txt` -- SAFE-fragment-encoded
|
| 18 |
+
model, for scaffold growth / local fragment editing.
|
| 19 |
+
|
| 20 |
+
Licensed under MIT -- free to use, modify, and redistribute.
|
| 21 |
+
|
| 22 |
+
Note: the training corpus includes data from ChEMBL, which is distributed
|
| 23 |
+
under CC-BY-SA. Whether that imposes any share-alike obligation on model
|
| 24 |
+
weights trained on it is an unsettled question in ML licensing generally --
|
| 25 |
+
not resolved here, mentioned for transparency.
|