BERT_SMILES_LARGE

This model is a 83.5M parameter ROBERTA model fine tuned on a dataset of 1.1M SMILES (Simplified molecular-input line-entry system) for masked language modeling (MLM). This model builds on BERT_SMILES which was fine tuned on only 50k SMILES.

Evaluation Loss: 0.482

Example:

Morphine

CN1CC[C@]23[C@@H]4[C@H]1CC5=C2C(=C(C=C5)O)O[C@H]3[C@H](C=C4)O

Intended uses & limitations

This model can now be used to predict physical or chemical properties with further training.

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 2.1.0+cu121
• Tokenizers 0.15.0