Views
No views yet
1from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
2
3tokenizer = AutoTokenizer.from_pretrained("sagawa/ReactionT5v2-retrosynthesis-USPTO_50k", return_tensors="pt")
4model = AutoModelForSeq2SeqLM.from_pretrained("sagawa/ReactionT5v2-retrosynthesis-USPTO_50k")
5
6inp = tokenizer('CCN(CC)CCNC(=S)NC1CCCc2cc(C)cnc21', return_tensors='pt')
7output = model.generate(**inp, num_beams=1, num_return_sequences=1, return_dict_in_generate=True, output_scores=True)
8output = tokenizer.decode(output['sequences'][0], skip_special_tokens=True).replace(' ', '').rstrip('.')
9output # 'CCN(CC)CCN=C=S.Cc1cnc2c(c1)CCCC2N'1cd task_retrosynthesis
2python finetune.py \
3 --output_dir='t5' \
4 --epochs=20 \
5 --lr=2e-5 \
6 --batch_size=32 \
7 --input_max_len=150 \
8 --target_max_len=150 \
9 --weight_decay=0.01 \
10 --evaluation_strategy='epoch' \
11 --save_strategy='epoch' \
12 --logging_strategy='epoch' \
13 --save_total_limit=10 \
14 --train_data_path='../data/USPTO_50k/train.csv' \
15 --valid_data_path='../data/USPTO_50k/val.csv' \
16 --disable_tqdm \
17 --model_name_or_path='sagawa/ReactionT5v2-retrosynthesis'| Model | Training set | Test set | Top-1 [% acc.] | Top-2 [% acc.] | Top-3 [% acc.] | Top-5 [% acc.] |
|---|---|---|---|---|---|---|
| Sequence-to-sequence | USPTO_50k | USPTO_50k | 37.4 | - | 52.4 | 57.0 |
| Molecular Transformer | USPTO_50k | USPTO_50k | 43.5 | - | 60.5 | - |
| SCROP | USPTO_50k | USPTO_50k | 43.7 | - | 60.0 | 65.2 |
| T5Chem | USPTO_50k | USPTO_50k | 46.5 | - | 64.4 | 70.5 |
| CompoundT5 | USPTO_50k | USPTO_50k | 44,2 | 55.2 | 61.4 | 67.3 |
| ReactionT5 | - | USPTO_50k | 13.8 | 18.6 | 21.4 | 26.2 |
| ReactionT5 (This model) | USPTO_50k | USPTO_50k | 71.2 | 81.4 | 84.9 | 88.2 |
@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}
}