| --- |
| license: mit |
| datasets: |
| - sagawa/ZINC-canonicalized |
| metrics: |
| - accuracy |
| model-index: |
| - name: ZINC-deberta |
| results: |
| - task: |
| name: Masked Language Modeling |
| type: fill-mask |
| dataset: |
| name: sagawa/ZINC-canonicalized |
| type: sagawa/ZINC-canonicalized |
| metrics: |
| - name: Accuracy |
| type: accuracy |
| value: 0.9497212171554565 |
| --- |
| |
| # CompoundT5 |
|
|
| This model is a re-pretrained version of [google/t5-v1_1-base](https://huggingface.co/microsoft/deberta-base) on the sagawa/ZINC-canonicalized dataset. |
| It achieves the following results on the evaluation set: |
| - Loss: 0.1202 |
| - Accuracy: 0.9497 |
|
|
|
|
| ## Model description |
|
|
| We trained t5 on SMILES from ZINC using masked-language modeling (MLM). Its tokenizer is also trained on ZINC. |
|
|
|
|
| ## Intended uses & limitations |
|
|
| This model can be used to predict molecules' properties, reactions, or interactions with proteins by changing the way of finetuning. |
| As an example, We finetuned this model to predict products. The model is [here](https://huggingface.co/sagawa/ZINC-t5-productpredicition), and you can use the demo [here](https://huggingface.co/spaces/sagawa/predictproduct-t5). |
| Using its encoder, we trained a regression model to predict a reaction yield. You can use this demo [here](https://huggingface.co/spaces/sagawa/predictyield-t5). |
|
|
| ## Training and evaluation data |
|
|
| We downloaded [ZINC data](https://drive.google.com/drive/folders/1lSPCqh31zxTVEhuiPde7W3rZG8kPgp-z) and canonicalized them using RDKit. Then, we dropped duplicates. The total number of data is 22992522, and they were randomly split into train:validation=10:1. |
|
|
| ## Training procedure |
|
|
| ### Training hyperparameters |
|
|
| The following hyperparameters were used during training: |
| - learning_rate: 5e-03 |
| - train_batch_size: 30 |
| - eval_batch_size: 32 |
| - seed: 42 |
| - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
| - lr_scheduler_type: linear |
| - num_epochs: 30.0 |
|
|
| ### Training results |
|
|
| | Training Loss | Step | Accuracy | Validation Loss | |
| |:-------------:|:------:|:--------:|:---------------:| |
| | 0.2471 | 25000 | 0.9843 | 0.2226 | |
| | 0.1871 | 50000 | 0.9314 | 0.1783 | |
| | 0.1791 | 75000 | 0.9371 | 0.1619 | |
| | 0.1596 | 100000 | 0.9401 | 0.1520 | |
| | 0.1522 | 125000 | 0.9422 | 0.1449 | |
| | 0.1435 | 150000 | 0.9436 | 0.1404 | |
| | 0.1421 | 175000 | 0.9447 | 0.1368 | |
| | 0.1398 | 200000 | 0.9459 | 0.1322 | |
| | 0.1297 | 225000 | 0.9466 | 0.1299 | |
| | 0.1324 | 250000 | 0.9473 | 0.1268 | |
| | 0.1257 | 275000 | 0.9483 | 0.1244 | |
| | 0.1266 | 300000 | 0.9491 | 0.1216 | |
| | 0.1301 | 325000 | 0.9497 | 0.1204 | |
|
|
|
|
| ## Citation |
|
|
| <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. --> |
| ``` |
| @article{Sagawa2025, |
| title = {ReactionT5: a pre-trained transformer model for accurate chemical reaction prediction with limited data}, |
| author = {Sagawa, Tatsuya and Kojima, Ryosuke}, |
| journal = {Journal of Cheminformatics}, |
| year = {2025}, |
| volume = {17}, |
| number = {1}, |
| pages = {126}, |
| doi = {10.1186/s13321-025-01075-4}, |
| url = {https://doi.org/10.1186/s13321-025-01075-4} |
| } |
| ``` |
