File size: 2,802 Bytes
75b2631 7298339 9bfe9f0 ceba297 dbbdbcb a56104d 9bfe9f0 75b2631 089d787 63cb0f6 089d787 51cb62c 089d787 17d47f2 63cb0f6 089d787 17d47f2 2a45f64 17d47f2 89aaa46 2a45f64 17d47f2 89aaa46 4b74b2f dbfa68a 5d63c09 dbfa68a 89aaa46 df1516e 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 50 51 52 53 54 55 56 57 58 59 60 61 |
---
tags:
- molecular language model
- SELFIES
- molecule generation
widget:
- text: "[C][=C][C][=C][C][=C][Ring1][=Branch1]"
inference:
parameters:
repetition_penalty: 100
num_return_sequences: 5
---
# 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, AutoModelForSeq2SeqLM
>>> tokenizer = AutoTokenizer.from_pretrained("zjunlp/MolGen")
>>> model = AutoModelForSeq2SeqLMq.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=15,
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][=C][C][=C]',
'[C][=C][C][=C][C][=C][Ring1][=Branch1][C@H1][C][=C][C]',
'[C][=C][C][=C][C][=C][Ring1][=Branch1][C@H1][=C][C][=C]']
```
### 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}
}
``` |