Instructions to use zhangtaolab/PlantHelixSeek-sequence_conservation with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use zhangtaolab/PlantHelixSeek-sequence_conservation with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-classification", model="zhangtaolab/PlantHelixSeek-sequence_conservation", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoModelForSequenceClassification model = AutoModelForSequenceClassification.from_pretrained("zhangtaolab/PlantHelixSeek-sequence_conservation", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
PlantHelixSeek-sequence_conservation
Fine-tuned checkpoint of PlantHelixSeek (470M parameters) for binary classification of conserved sequence.
Task
- Type: binary classification of conserved sequence
- Label: 0 = non-conserved, 1 = conserved
- Input length: 1,000 bp
- Dataset: plant-multi-species-sequence-conservation
Dataset and splits
The checkpoint uses the released train.csv, dev.csv and test.csv splits of plant-multi-species-sequence-conservation (56,320 / 7,040 / 7,040 sequences). The source of the dataset is described on its dataset card.
Training
| Setting | Value |
|---|---|
| Base model | PlantHelixSeek-470M |
| Epochs | 3 |
| Learning rate | 2e-5 |
| Per-device batch size | 1 |
| Gradient accumulation | 16 |
| Effective batch size | 16 |
| Warm-up ratio | 0.1 |
| Weight decay | 0.01 |
| Scheduler | linear |
| Optimizer | AdamW (fused) |
| Precision | bf16 |
| Gradient clipping | 1.0 |
| Seed | 42 |
| Hardware | one NVIDIA GPU |
| Training loss | 0.203 |
| Training time | 1.9 h |
Held-out validation metrics
Metrics on the held-out dev.csv split (the split used for checkpoint selection).
| Metric | Value |
|---|---|
| f1 | 0.914 |
| AUROC | 0.969 |
| AUPRC | 0.970 |
| mcc | 0.828 |
| accuracy | 0.914 |
Usage
from transformers import AutoModelForSequenceClassification, AutoTokenizer
model = AutoModelForSequenceClassification.from_pretrained(
"zhangtaolab/PlantHelixSeek-sequence_conservation", trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained(
"zhangtaolab/PlantHelixSeek-sequence_conservation", trust_remote_code=True
)
Always pass trust_remote_code=True; the HelixSeek architecture is registered through the Hugging Face auto_map mechanism.
Citation
Wu Y, He Y, Chen L, Liu Y, Wang T, Zhao H, Zhang Y, Zhang T. PlantHelixSeek learns transferable genome representations that capture conserved genome grammar for functional interpretation across plant species. Molecular Plant, 2026. Manuscript MOLECULAR-PLANT-D-26-00999 (DOI to be assigned).
Dataset source: Liu et al., Molecular Plant 2025.
License
CC-BY-NC 4.0 (research use; contact the authors for commercial licensing).
- Downloads last month
- 33
Model tree for zhangtaolab/PlantHelixSeek-sequence_conservation
Base model
zhangtaolab/PlantHelixSeek