CreatiPhage-145M

Autoregressive DNA language model for phage sequence generation, and the pretrained backbone of CPDD.

Code · Technical report · CPDD-147M

144.5M parameters · 98,304-token context · FP32 safetensors

Quickstart

Requires the accompanying PyTorch code.

git clone https://github.com/dhnihaoya/CreatiPhage-Diffusion-DNA-Lm.git
cd CreatiPhage-Diffusion-DNA-Lm
pip install -r requirements.txt
python examples/generate_ar.py \
  --checkpoint dinghhhhhhhhhhhhhhh/CreatiPhage-145M \
  --primer ATGCGT --length 16 --device cpu

General six-epoch checkpoint, not a species-specific fine-tune. Generated sequences require independent evaluation.

Weights: CC BY 4.0, with attribution to CreatiPhage Biotechnology Co., Ltd. Code: Apache-2.0.