blenderwang commited on
Commit
565aa42
·
verified ·
1 Parent(s): 156de07

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +1 -7
README.md CHANGED
@@ -1,24 +1,18 @@
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
 
1
  ---
2
  license: mit
 
 
3
  ---
4
 
5
  # Mol ID
6
 
7
  A transformer encoder model pretrained on 50M ZINC SMILES string using flash attention 2
8
 
 
 
9
  Hardware:
10
+ - gpu that support flash attention 2 and bf16
 
11
 
12
  Software:
13
  - flash attention 2
14
  - lightning for mixed precision (bf16-mixed)
15
  - wandb for logging
 
16
  - huggingface
17
  - tokenizers
18
  - datasets