Commit
•
909cce2
1
Parent(s):
3e98081
Update README.md
Browse files
README.md
CHANGED
|
@@ -28,4 +28,21 @@ configs:
|
|
| 28 |
path: data/val-*
|
| 29 |
- split: test
|
| 30 |
path: data/test-*
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
path: data/val-*
|
| 29 |
- split: test
|
| 30 |
path: data/test-*
|
| 31 |
+
license: mit
|
| 32 |
+
task_categories:
|
| 33 |
+
- token-classification
|
| 34 |
+
tags:
|
| 35 |
+
- chemistry
|
| 36 |
+
- biology
|
| 37 |
+
- medical
|
| 38 |
+
size_categories:
|
| 39 |
+
- 1B<n<10B
|
| 40 |
---
|
| 41 |
+
|
| 42 |
+
[ZINC20](https://zinc20.docking.org/) Dataset with [SELFIES](https://arxiv.org/abs/1905.13741) added. Any smile that could not be successfully converted was dropped from the dataset.
|
| 43 |
+
|
| 44 |
+
Every tranch was downloaded, this is not the ~1B example ML subset from https://files.docking.org/zinc20-ML/.
|
| 45 |
+
|
| 46 |
+
The dataset was entirely shuffled then split into 80%/10%/10% splits for train/val/test.
|
| 47 |
+
|
| 48 |
+
A file vocab.json is in the root of the reposity that contains all of the SELFIES tokens found in the data, with [START], [STOP], and [PAD] added.
|