Download of several databases when model type just esm1b

[oliverdutton]

I ran

export NEW_PROT_FASTA_FILE_PATH='/raid/app/oliver/tmp/disprot_2022_03.fasta' 
export NEW_PROT_RES_REPR_DIR_PATH='/raid/app/oliver/tmp/'

then the script (last bit of https://github.com/PeptoneInc/ADOPT/blob/main/scripts/adopt_inference.sh)

TRAIN_STRATEGY="train_on_cleared_1325_test_on_117_residue_split"
MODEL_TYPE="esm-1b"
PRED_Z_FILE_PATH="predicted_z_scores.json"

printf "Extracting residue level representation of %s \n" $NEW_PROT_FASTA_FILE_PATH

python adopt/embedding.py $NEW_PROT_FASTA_FILE_PATH \
                          $NEW_PROT_RES_REPR_DIR_PATH

printf "Computing Z scores of %s \n" $NEW_PROT_FASTA_FILE_PATH

python adopt/inference.py $NEW_PROT_FASTA_FILE_PATH \
                          $NEW_PROT_RES_REPR_DIR_PATH \
                          $PRED_Z_FILE_PATH \
                          --train_strategy $TRAIN_STRATEGY \
                          --model_type $MODEL_TYPE \
                          --mode "bulk"

on ghcr.io/peptoneinc/adopt:latest

In the output it says:

Downloading: "https://dl.fbaipublicfiles.com/fair-esm/models/esm1v_t33_650M_UR90S_1.pt" to /root/.cache/torch/hub/checkpoints/esm1v_t33_650M_UR90S_1.pt

then

Downloading: "https://dl.fbaipublicfiles.com/fair-esm/models/esm1b_t33_650M_UR50S.pt" to /root/.cache/torch/hub/checkpoints/esm1b_t33_650M_UR50S.pt

then

Downloading: "https://dl.fbaipublicfiles.com/fair-esm/regression/esm1b_t33_650M_UR50S-contact-regression.pt" to /root/.cache/torch/hub/checkpoints/esm1b_t33_650M_UR50S-contact-regression.pt

I said esm-1b model type but esm-1v are both being download and predicted with and then contact regression too. This is either a bug, or confusingly named.

[CFisicaro]

The docker container is currently not constantly up to date with the repository, please clone and install the library from scratch every time for now.

Could you check if the error still persists?