|
--- |
|
license: mit |
|
tags: |
|
- dna |
|
- language-model |
|
- variant-effect-prediction |
|
- biology |
|
- genomics |
|
datasets: |
|
- songlab/genomes-brassicales-balanced-v1 |
|
--- |
|
# GPN trained on *Arabidopsis thaliana* and 7 other Brassicales |
|
See https://github.com/songlab-cal/gpn for more details. |
|
|
|
Basic usage: |
|
```python |
|
import gpn.model |
|
from transformers import AutoModel, AutoModelForMaskedLM |
|
|
|
model = AutoModel.from_pretrained("songlab/gpn-brassicales") |
|
# or |
|
model = AutoModelForMaskedLM.from_pretrained("songlab/gpn-brassicales") |
|
``` |
|
|
|
## Some hparams: |
|
* repeat_weight: 0.1 |
|
* lr: 120k at 1e-3 + 30k cosine decay |
