File size: 2,736 Bytes
75b2631
 
 
 
7298339
75b2631
089d787
 
63cb0f6
089d787
 
 
 
51cb62c
089d787
17d47f2
63cb0f6
089d787
17d47f2
 
 
89aaa46
17d47f2
89aaa46
a4a41af
17d47f2
89aaa46
 
4b74b2f
dbfa68a
 
 
 
 
 
89aaa46
 
17d47f2
 
 
089d787
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
---
tags:
- molecular language model
- SELFIES
- molecule generation
---
# MolGen
MolGen was introduced in the paper ["Molecular Language Model as Multi-task Generator"](https://arxiv.org/pdf/2301.11259.pdf) and first released in [this repository](https://github.com/zjunlp/MolGen). It is a pre-trained molecular generative model built using the 100\% robust molecular language representation, SELFIES.

## Model description
MolGen is the first pre-trained model that only produces chemically valid molecules. 
With a training corpus of over 100 million molecules in SELFIES representation, MolGen learns the intrinsic structural patterns of molecules by mapping corrupted SELFIES to their original forms.
Specifically, MolGen employs a bidirectional Transformer as its encoder and an autoregressive Transformer as its decoder.
Through its carefully designed multi-task molecular prefix tuning (MPT), MolGen can generate molecules with desired properties, making it a valuable tool for molecular optimization.

## Intended uses
You can use the raw model for molecular generation or fine-tune it to a downstream task. See the [repository](https://github.com/zjunlp/MolGen) to look for fine-tune details on a task that interests you.

### How to use
Molecule generation example:
```python
>>> from transformers import AutoTokenizer, BartForConditionalGeneration

>>> tokenizer = AutoTokenizer.from_pretrained("zjunlp/MolGen")
>>> model = BartForConditionalGeneration.from_pretrained("zjunlp/MolGen")

>>> sf_input = tokenizer("[C][=C][C][=C][C][=C][Ring1][=Branch1]", return_tensors="pt")
>>> # beam search
>>> molecules = model.generate(input_ids=sf_input["input_ids"],
                              attention_mask=sf_input["attention_mask"],
                              max_length=20,
                              min_length=5,
                              num_return_sequences=5,
                              num_beams=5,
                              past_prompt=None)
>>> sf_output = [tokenizer.decode(g, skip_special_tokens=True, clean_up_tokenization_spaces=True).replace(" ","") for g in molecules]
['[C][=C][C][=C][C][=C][Ring1][=Branch1]', '[C][=C][C][=C][C][=C][C][=C][Ring1][=Branch1]', '[C][=C][C][=C][C][=C][Ring1][=Branch1][C@H1][C][=C][C][=C][C][=C][Ring1][=Branch1]', '[C][=C][C][=C][C][=C][Ring1][=Branch1][C][=C][C][=C][C][=C][Ring1][=Branch1]', '[C][=C][C][=C][C][=C][Ring1][=Branch1][C@H1][=C][C][=C][Ring1][=Branch1]']
```


### BibTeX entry and citation info
```bibtex
@article{fang2023molecular,
  title={Molecular Language Model as Multi-task Generator},
  author={Fang, Yin and Zhang, Ningyu and Chen, Zhuo and Fan, Xiaohui and Chen, Huajun},
  journal={arXiv preprint arXiv:2301.11259},
  year={2023}
}
```