molcrawl-compounds-gpt2-medium

Model Description

GPT-2 medium (345M parameters) foundation model pre-trained on compound SMILES strings from the MolCrawl dataset.

The tokenizer is a character-level BPE tokenizer (vocab_size=612) that encodes each SMILES character as a separate token. Input SMILES strings should be passed without spaces (e.g. CC(=O)O). The [SEP] token (id=13) is used as the end-of-sequence marker.

Datasets

• MolCrawl compounds corpus (chembl + zinc + opv + reddb + pubchemqc): https://github.com/mmai-framework-lab/MolCrawl-HFuploader/blob/main/workflows/hugging_face/run_upload_hf.sh (Pre-training corpus)

• Model Type: gpt2

• Data Type: Molecule/Compound

• Training Date: 2026-04-15

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = AutoModelForCausalLM.from_pretrained("kojima-lab/molcrawl-compounds-gpt2-medium")
tokenizer = AutoTokenizer.from_pretrained("kojima-lab/molcrawl-compounds-gpt2-medium")

# Generate SMILES string
prompt = "CC(=O)O"
input_ids = tokenizer.encode(prompt, return_tensors="pt")
with torch.no_grad():
    output_ids = model.generate(
        input_ids,
        max_new_tokens=50,
        do_sample=True,
        temperature=0.8,
        eos_token_id=tokenizer.convert_tokens_to_ids("[SEP]"),  # [SEP] is EOS for compounds
        pad_token_id=0,
    )
print(tokenizer.decode(output_ids[0], skip_special_tokens=True))

Training

This model was trained with the RIKEN Foundation Model pipeline. For more details, please refer to the training configuration files included in this repository.

License

This model is released under the APACHE-2.0 license.

Citation

If you use this model, please cite:

@misc{molcrawl_compounds_gpt2_medium,
  title={molcrawl-compounds-gpt2-medium},
  author={{RIKEN}},
  year={2026},
  publisher={{Hugging Face}},
  url={{https://huggingface.co/kojima-lab/molcrawl-compounds-gpt2-medium}}
}