vigg_ribonanza

Source code of the Stanford Ribonanza RNA Folding first place solution

You can read the detailed solution description here -- https://www.kaggle.com/competitions/stanford-ribonanza-rna-folding/discussion/460121

HARDWARE

Ubuntu 20.04.4 LTS x86_64

CPU: 2x EPYC 7662 (128 CPU each)

6 GPU NVIDIA Tesla V100 32GB CoWoS HBM2

1000GB RAM

2TB SSD

Installation steps

Use setup_data.sh to download train and test data

Use setup_tools.sh to install all required software

Use calculate_features.sh to calculate features required for the model

How to train your model

To train model use python_scripts/train_uni_adjnet.py script

We recommend to train model with the following options

python3 python_scripts/train_uni_adjnet_se.py --bpp_path /projects_nvme/deepbeer/eterna/ --train_path /projects/deepbeer/ribonanza/train_data/train_data.parquet  --out_path  outmodel_dir --device 0 --num_workers 20 --wd 0.05 --epoch 270 --lr_max 5e-3 --pct_start 0.05 --batch_cnt 1791 --sgd_lr 5e-5 --sgd_epochs 25 --sgd_batch_cnt 500 --sgd_wd 0.05 --fold 0 --nfolds 1000 --pos_embedding dyn --adj_ks 3 --seed 42 --use_se

How to make predictions on a new test set.

Use python_scripts/train_uni_adjnet.py

Example:

python3 python_scripts/predict.py --bpp_path $bpp_path  --test_path $test_path --model_path $model_weights_path --out_path $out_dir --device 0 --pos_embedding dyn --adj_ks 3 --num_workers 20 --use_se

Contacts

In case of any question you write to [email protected]