QUESTION: is it ok if I keep getting the message: Could not load pretrained MHT weights, skipping

[DanLep97]

• PIGT version:
• Python version: 3.10 (as in environment.yml)
• Operating System: Ubuntu 22

Description

I'm getting predictions without the pre-trained MHT weights loaded. I downloaded the weights, adding the path as an argument but I keep getting this warning.

Should I be worried about this?

What I Did

rm -rf ../../databases/NeuralPLexer/inferences/*

declare -a RECEPTORS=(`cat rec_paths.txt`)
declare -a LIGANDS=(`cat lig_paths.txt`)
declare -a OUTPUTS=(`cat output_paths.txt`)

len=${#OUTPUTS[@]}
for (( i=0; i<${len}; i++));
do
        printf "Current index %d with value %s\n" $i "${OUTPUTS[$i]}"
        mkdir "${OUTPUTS[$i]}"

        neuralplexer-inference --task=batched_structure_sampling \
                --input-receptor "${RECEPTORS[$i]}" \
                --input-ligand "${LIGANDS[$i]}" \
                --use-template \
                --input-template "${RECEPTORS[$i]}" \
                --out-path "${OUTPUTS[$i]}" \
                --model-checkpoint /home/daniil/Desktop/NeuralPLexer/models/complex_structure_prediction.ckpt \
                --n-samples 10 \
                --chunk-size 1 \
                --num-steps=40 \
                --cuda \
                --sampler=langevin_simulated_annealing\
                --rank-outputs-by-confidence
done