SNS Paper - Borzoi Fine-tuned Models
Collection
Pre-trained Borzoi foundation model (johahi/borzoi-replicate-0) fine-tuned on mouse ESC-derived CUT&Tag H3K27me3 and H3K4me3 tracks at 32bp resolution • 10 items • Updated
Borzoi Fine-tuned RF 1k — Foundation Model
Pre-trained Borzoi model (johahi/borzoi-replicate-0) fine-tuned on mouse ESC-derived CUT&Tag H3K27me3 and H3K4me3 tracks via two-stage training (linear probe then full fine-tuning).
- Receptive field: 1k
- Base model:
johahi/borzoi-replicate-0 - Resolution: 32 bp
- Training data: Normalized CUT&Tag from mouse ESC-derived embryoid bodies
- Paper: Casimov, Lifshitz et al., "Reverse engineering the genomic encodings of the pluripotent epigenome"
Usage
torchrun --nproc_per_node=8 infer_borzoi_pytorch.py \
--config config.yaml \
--checkpoint model.safetensors \
--genome_fasta mm10.fa \
--output_dir predictions/
See the paper companion repository for full code and configs.
- Downloads last month
- -
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support