RxnScribe - GitLite

KO | EN

gitlite — search

#javascript #hacktoberfest #typescript #llm #android #deep-learning #macos #nodejs #java #awesome #cli #kubernetes #pytorch #open-source

RxnScribe

(★ 117)

A Sequence Generation Model for Reaction Diagram Parsing

파일 탐색기

최종 버전 다운로드 (.zip)

- - output0.png
  - output1.png
- acs.joc.5b01703-Scheme-c1.png
- acs.oprd.5b00209-Scheme-c4.png
- acs.orglett.5b01044-Table-c1.png
- acs.orglett.6b01181-Scheme-c4.png
- jacs.5b12989-Table-c3.png
- model.png
- - dev.json
  - dois.txt
  - train.json
- predict.ipynb
- visualize_data.ipynb
- - __init__.py
  - beam_search.py
  - decode_strategy.py
  - greedy_search.py
- - __init__.py
  - attention_layer.py
  - backbone.py
  - misc.py
  - pix2seq.py
  - position_encoding.py
  - transformer.py
- - __init__.py
  - decoder.py
  - embedding.py
  - swin_transformer.py
- __init__.py
- data.py
- dataset.py
- evaluate.py
- interface.py
- loss.py
- model.py
- tokenizer.py
- transforms.py
- utils.py
- train_pix2seq_cv.sh
- train_pix2seq_full.sh
.gitignore
evaluate.py
LICENSE
main.py
predict.py
README.md
requirements.txt
setup.py

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# RxnScribe

This is the repository for RxnScribe, a sequence generation model for reaction diagram parsing.
Try our [demo](https://huggingface.co/spaces/yujieq/RxnScribe) on Hugging Face!

![](https://raw.githubusercontent.com/thomas0809/RxnScribe/main/assets/model.png)

If you use RxnScribe in your research, please cite our [paper](https://pubs.acs.org/doi/10.1021/acs.jcim.3c00439).
```
@article{
    RxnScribe,
    author = {Qian, Yujie and Guo, Jiang and Tu, Zhengkai and Coley, Connor W. and Barzilay, Regina},
    title = {RxnScribe: A Sequence Generation Model for Reaction Diagram Parsing},
    journal = {Journal of Chemical Information and Modeling},
    doi = {10.1021/acs.jcim.3c00439}
}
```

Molecule structure recognition is supported by MolScribe 
([paper](https://pubs.acs.org/doi/10.1021/acs.jcim.2c01480), 
[code](https://github.com/thomas0809/MolScribe), 
[demo](https://huggingface.co/spaces/yujieq/MolScribe)).

## Quick Start
Run the following command to install the package and its dependencies:
```
git clone git@github.com:thomas0809/RxnScribe.git
cd RxnScribe
python setup.py install
```

Download the checkpoint and use RxnScribe to extract reactions from a diagram:
```python
import torch
from rxnscribe import RxnScribe
from huggingface_hub import hf_hub_download

ckpt_path = hf_hub_download("yujieq/RxnScribe", "pix2seq_reaction_full.ckpt")
model = RxnScribe(ckpt_path, device=torch.device('cpu'))

image_file = "assets/jacs.5b12989-Table-c3.png"
predictions = model.predict_image_file(image_file, molscribe=True, ocr=True)
```
The predictions will be in the following format:
```python
[
    {  # First reaction
        'reactants': [
            {
                'category': '[Mol]', 'category_id': 1, 'bbox': (0.1550, 0.0246, 0.2851, 0.2614),
                'smiles': '*OC(=O)c1ccccc1C#Cc1ccccc1', 'molfile': '(omitted)' 
            }, 
            # ... more reactants 
        ],
        'conditions': [
            {
                'category': '[Txt]', 'category_id': 2, 'bbox': (0.2941, 0.0641, 0.3811, 0.1450),
                'text': ['CIBcat', '(1.4 equiv)']
            }, 
            # ... more conditions
        ],
        'products': [ 
            # ...
        ]
    },
    # More reactions
]
```
We provide a function to visualize the prediction:
```python
visualize_images = model.draw_predictions(predictions, image_file=image_file)
```
Each predicted reaction will be visualized in a separate image, where 
<b style="color:red">red boxes are <i><u style="color:red">reactants</u></i>,</b>
<b style="color:green">green boxes are <i><u style="color:green">reaction conditions</u></i>,</b>
<b style="color:blue">blue boxes are <i><u style="color:blue">products</u></i>.</b>

<img src="assets/output/output0.png" width="384"/> <img src="assets/output/output1.png" width="384"/>

This [notebook](notebook/predict.ipynb) shows how to run RxnScribe and visualize the prediction.

For development or reproducing the experiments, follow the instructions below.

## Requirements
Install the required packages
```
pip install -r requirements.txt
```

## Data
Download the reaction diagrams from this [link](https://huggingface.co/yujieq/RxnScribe/blob/main/images.zip), 
and save them to `data/parse/images/`.

The ground truth files can be found at [`data/parse/splits/`](data/parse/splits/).

We perform five-fold cross validation in our experiments. The train/dev/test split for each fold is available.

This [notebook](notebook/visualize_data.ipynb) shows how to visualize the diagram and the ground truth.

## Train and Evaluate RxnScribe
Run this script to train and evaluate RxnScribe with five-fold cross validation.
```bash
bash scripts/train_pix2seq_cv.sh
```
Finally, we train RxnScribe with 90% of the dataset, and use the remaining 10% as the dev set.
We have released the [model checkpoint](https://huggingface.co/yujieq/RxnScribe/blob/main/pix2seq_reaction_full.ckpt).
```bash
bash scripts/train_pix2seq_full.sh
```