genomenet/crispr-array-detection

1---2title: CRISPR Array Detection3emoji: 🧬4colorFrom: gray5colorTo: gray6sdk: docker7pinned: false8license: mit9short_description: Detect CRISPR arrays in DNA sequences10---11 12# crispr-detect13 14BERT-based CRISPR array detection in prokaryotic genomes.15 16## Model17 18| | |19|---|---|20| architecture | BERT, 24 layers, 768 hidden, 430M params |21| input | DNA sequence (min 1000 bp) |22| output | per-position probability (0-1) |23 24## Deployment25 26### Push changes27 28```bash29cd /vol/hpcprojects/pmuench/crispr_tool/crispr-hf-space30git add -A31git commit -m "description"32git push33```34 35### Git credentials (first time)36 37```bash38git config --global credential.helper store39huggingface-cli login40# paste token from https://huggingface.co/settings/tokens41```42 43### Clone fresh44 45```bash46git clone https://huggingface.co/spaces/genomenet/crispr-array-detection47```48 49### Space settings (HuggingFace web UI)50 51- SDK: Docker52- Hardware: CPU Basic works for the default demo; T4 GPU is recommended for long sequences or low stride values53- Visibility: Public54 55### Model weights56 57Hosted at: https://huggingface.co/genomenet/crispr-bert-model58 59Downloaded automatically via `huggingface_hub` at startup.60 61## Local dev62 63```bash64pip install -r requirements.txt65python app.py66# http://localhost:786067```68 69## Files70 71```72├── app.py              # gradio app73├── inference/74│   ├── model_loader.py # model download75│   ├── tokenizer.py    # sequence validation76│   └── inference.py    # prediction77├── Dockerfile78└── requirements.txt79```80 81## Acknowledgements82 83- Ziyu Mu (HZI BIFO)84- DFG SPP 2141 (MC 172)85- BMBF GenomeNet86