|
--- |
|
license: mit |
|
datasets: |
|
- sagawa/pubchem-10m-canonicalized |
|
metrics: |
|
- accuracy |
|
model-index: |
|
- name: PubChem-10m-t5 |
|
results: |
|
- task: |
|
name: Masked Language Modeling |
|
type: fill-mask |
|
dataset: |
|
name: sagawa/pubchem-10m-canonicalized |
|
type: sagawa/pubchem-10m-canonicalized |
|
metrics: |
|
- name: Accuracy |
|
type: accuracy |
|
value: 0.9189779162406921 |
|
--- |
|
|
|
# PubChem-10m-t5 |
|
|
|
This model is a fine-tuned version of [google/t5-v1_1-base](https://huggingface.co/microsoft/deberta-base) on the sagawa/pubchem-10m-canonicalized dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.2165 |
|
- Accuracy: 0.9190 |
|
|
|
|
|
## Model description |
|
|
|
We trained t5 on SMILES from PubChem using the task of masked-language modeling (MLM). Compared to PubChem-10m-t5, PubChem-10m-t5-v2 uses a character-level tokenizer, and it was also trained on PubChem. |
|
|
|
|
|
## Intended uses & limitations |
|
|
|
This model can be used for the prediction of molecules' properties, reactions, or interactions with proteins by changing the way of finetuning. |
|
|
|
## Training and evaluation data |
|
|
|
We downloaded [PubChem data](https://drive.google.com/file/d/1ygYs8dy1-vxD1Vx6Ux7ftrXwZctFjpV3/view) and canonicalized them using RDKit. Then, we dropped duplicates. The total number of data is 9999960, 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 |
|
- seed: 42 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
- num_epochs: 10.0 |
|
|
|
### Training results |
|
|
|
| Training Loss | Step | Accuracy | Validation Loss | |
|
|:-------------:|:------:|:--------:|:---------------:| |
|
| 0.2592 | 100000 | 0.8997 | 0.2784 | |
|
| 0.2790 | 200000 | 0.9095 | 0.2468 | |
|
| 0.2278 | 300000 | 0.9162 | 0.2256 | |
