RNADiffusion: Generative RNA Secondary Structure Prediction using DiT

1.Please use the .yml file to create your environment

conda env create -f requirements.yml

2. During training, the hyperparameters are set as follows

python train.py --device cuda:0 \
                --diffusion_dim 8 \ 
                --diffusion_steps 20 \
                --cond_dim 8 \
                --batch_size 1 \
                --dp_rate 0.1 \
                --lr 0.0001 \
                --warmup 5 \
                --seed 2024 \
                --log_wandb False \
                --epochs 50 \
                --eval_every 3 \
                --u_conditioner_ckpt ufold_train_alldata.pt \
                --esm_conditioner_ckpt esm.ckpt \

3. Data

python data/binning_data.py
python data/batching_data.py
python data/get_base_info.py