Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
datasets:
|
| 4 |
+
- blenderwang/zinc-50M
|
| 5 |
+
---
|
| 6 |
+
|
| 7 |
+
# Mol ID
|
| 8 |
+
|
| 9 |
+
A transformer encoder model pretrained on 50M ZINC SMILES string using flash attention 2
|
| 10 |
+
|
| 11 |
+
With modern hardware and software stacks, it only took me 6~7 hours to pretrain this 17M model on 50M molecules for 5 epochs
|
| 12 |
+
|
| 13 |
+
Hardware:
|
| 14 |
+
- 4 cores cpu
|
| 15 |
+
- 1 RTX 3090 gpu
|
| 16 |
+
|
| 17 |
+
Software:
|
| 18 |
+
- flash attention 2
|
| 19 |
+
- lightning for mixed precision (bf16-mixed)
|
| 20 |
+
- wandb for logging
|
| 21 |
+
- [report](https://api.wandb.ai/links/blenderwang/5qou429x)
|
| 22 |
+
- huggingface
|
| 23 |
+
- tokenizers
|
| 24 |
+
- datasets
|
| 25 |
+
|
| 26 |
+
|
| 27 |
+
|
| 28 |
+
|
| 29 |
+
github repo: [link](https://github.com/BlenderWang9487/mol_id)
|