additional changes to bash

bash_files/fseek_after_human_model_dscript_cmap_ot/results.log

@@ -1,6 +0,0 @@
-[2022-12-26-12:51:38] D-SCRIPT Version 0.2.2
-[2022-12-26-12:51:38] Called as: /scratch2/kdevko01/conda/.conda/envs/dscript/bin/dscript train --train seqs-pairs/pairs/human_train.tsv --test seqs-pairs/pairs/human_test.tsv --embedding embeddings/human.h5 -o fseek_after_human_model_dscript_cmap_ot/results.log --save-prefix fseek_after_human_model_dscript_cmap_ot/ep_ --lr 0.0005 --lambda 0.05 --num-epoch 10 --weight-decay 0 --batch-size 25 --pool-width 9 --kernel-width 7 --dropout-p 0.2 --projection-dim 100 --hidden-dim 50 --kernel-width 7 --device 3 --run-cmap --contact-map-train ../data/pairs/cmap_train_lt_400.tsv --contact-map-test ../data/pairs/cmap_test_lt_400.tsv --contact-map-mode ot --contact-map-embedding ../lynnfiles/new_cmap_embed --contact-maps ../data/embeddings/cmap_filtered_lt_400.h5 --contact-map-sampler ../data/models/sampler/iter_9.sav --ot-cmap-nsamples 100 --contact-map-lr 1 --contact-map-lambda 0.2
-[2022-12-26-12:51:38] Using CUDA device 3 - Tesla V100-PCIE-32GB
-[2022-12-26-12:51:39] Loaded 843584 training pairs
-[2022-12-26-12:51:39] Loaded 52725 test pairs
-[2022-12-26-12:51:39] Loading embeddings...

bash_files/test.sh

@@ -1,6 +1,6 @@
 #!/bin/bash
 
-ORGS=( ecoli yeast mouse fly )
+ORGS=( fly human mouse ) # ecoli yeast
 
 TOPSY_TURVY=
 EMBEDDING_DIR=embeddings/
@@ -51,4 +51,4 @@ done
 
 #Topsyturvy BEFORE: ./test.sh -d 2 -m fseek_before_human_model_dscript/ep__epoch01.sav -T fseek_before -D fseek_before_human_model_dscript/eval -f results -t
 
-#Topsyturvy DSCRIPT: ./test.sh -d 3 -m original_human_model_dscript/ep__epoch03.sav -T dscript -D original_human_model_dscript/eval -f results -t
+#Topsyturvy DSCRIPT: ./test.sh -d 5 -m fseek_after_human_model_topsyturvy_cmap_ot/ep__epoch02.sav -T dscript -D fseek_after_human_model_topsyturvy_cmap_ot/eval -f results -t

bash_files/train_foldseek_after-cmap-ot.sh

@@ -24,7 +24,7 @@ while getopts "d:t:T:e:vo:p:s:c:C:l:m:" args; do
         ;;
         e) EMBEDDING=${OPTARG}
         ;;
-        v) TOPSY_TURVY="--topsy-turvy --glider-weight 0.2 --glider-thres 0.925"; OUTPUT_FOLDER=fseek_after_human_model_topsyturvy;
+        v) TOPSY_TURVY="--topsy-turvy --glider-weight 0.2 --glider-thres 0.925"; OUTPUT_FOLDER=fseek_after_human_model_topsyturvy_cmap_ot;
         ;;
         o) OUTPUT_FOLDER=${OPTARG}
         ;;
@@ -45,7 +45,7 @@ done
     
 if [ ! -d ${OUTPUT_FOLDER} ]; then mkdir -p $OUTPUT_FOLDER; fi
 
-
+#./train_foldseek_after-cmap-ot.sh -v -s ../data/models/sampler/sampler-run-Mon-26-Dec-2022-12\:07\:02-PM-EST/iter_999.sav -d 3
 
 dscript train --train $TRAIN --test $TEST --embedding $EMBEDDING $TOPSY_TURVY \
               -o ${OUTPUT_FOLDER}/results.log \
@@ -53,4 +53,4 @@ dscript train --train $TRAIN --test $TEST --embedding $EMBEDDING $TOPSY_TURVY \
               --lr 0.0005 --lambda 0.05 --num-epoch 10 \
               --weight-decay 0 --batch-size 25 --pool-width 9 \
               --kernel-width 7 --dropout-p 0.2 --projection-dim 100 \
-              --hidden-dim 50 --kernel-width 7 --device $DEVICE --run-cmap --contact-map-train ${CMAP_TRAIN} --contact-map-test ${CMAP_TEST} --contact-map-mode ot --contact-map-embedding ${CMAP_LANG_EMB} --contact-maps ${CMAP_EMB} --contact-map-sampler ${SAMPLER} --ot-cmap-nsamples 100 --contact-map-lr 1 --contact-map-lambda 0.2 # --allow_foldseek --foldseek_fasta ${FOLDSEEK_FASTA} --foldseek_vocab ${FOLDSEEK_VOCAB} --add_foldseek_after_projection  ## need to add the foldseek part
+              --hidden-dim 50 --kernel-width 7 --device $DEVICE --run-cmap --contact-map-train ${CMAP_TRAIN} --contact-map-test ${CMAP_TEST} --contact-map-mode ot --contact-map-embedding ${CMAP_LANG_EMB} --contact-maps ${CMAP_EMB} --contact-map-sampler ${SAMPLER} --ot-cmap-nsamples 100 --contact-map-lr 0.0001 --contact-map-lambda 0.1 # --allow_foldseek --foldseek_fasta ${FOLDSEEK_FASTA} --foldseek_vocab ${FOLDSEEK_VOCAB} --add_foldseek_after_projection  ## need to add the foldseek part

bash_files/train_original.sh

@@ -4,7 +4,7 @@ TOPSY_TURVY=
 TRAIN=seqs-pairs/pairs/human_train.tsv
 TEST=seqs-pairs/pairs/human_test.tsv
 EMBEDDING=embeddings/human.h5
-OUTPUT_FOLDER=original_human_model_dscript
+OUTPUT_FOLDER=original_human_model_dscript-v2
 OUTPUT_PREFIX=results-
 while getopts "d:t:T:e:vo:p:" args; do
     case $args in
@@ -16,7 +16,7 @@ while getopts "d:t:T:e:vo:p:" args; do
         ;;
         e) EMBEDDING=${OPTARG}
         ;;
-        v) TOPSY_TURVY="--topsy-turvy --glider-weight 0.2 --glider-thres 0.925"
+        v) TOPSY_TURVY="--topsy-turvy --glider-weight 0.2 --glider-thres 0.925"; OUTPUT_FOLDER=original_human_model_tt-v2
         ;;
         o) OUTPUT_FOLDER=${OPTARG}
         ;;
@@ -30,9 +30,9 @@ if [ ! -d ${OUTPUT_FOLDER} ]; then mkdir -p $OUTPUT_FOLDER; fi
 
 
 dscript train --train $TRAIN --test $TEST --embedding $EMBEDDING $TOPSY_TURVY \
-              --o ${OUTPUT_FOLDER}/results.log \
+              -o ${OUTPUT_FOLDER}/results.log \
               --save-prefix  ${OUTPUT_FOLDER}/ep_ \
               --lr 0.001 --lambda 0.05 --num-epoch 10 \
               --weight-decay 0 --batch-size 25 --pool-width 9 \
               --kernel-width 7 --dropout-p 0.2 --projection-dim 100 \
-              --hidden-dim 50 --kernel-width 7 --device $DEVICE
+              --hidden-dim 50 --kernel-width 7 --device $DEVICE

bash_files/train_sampler.sh

@@ -35,4 +35,4 @@ echo "MAX_DATA=${MAX_DATA}, ITER=${ITER}, DEVICE=${DEVICE}, OUTPUT=${OUTPUT}, LR
 
 if [ ! -d $OUTPUT ]; then mkdir $OUTPUT; fi
 
-dscript sampler  --embedding $EMBEDDING --output $OUTPUT --device $DEVICE $CHECKPOINT --lr $LR --max-data ${MAX_DATA} --save-at-iter ${SAVE_AT_ITER}
+dscript sampler  --embedding $EMBEDDING --output $OUTPUT --device $DEVICE $CHECKPOINT --iter $ITER --lr $LR --max-data ${MAX_DATA} --save-at-iter ${SAVE_AT_ITER}

dscript/commands/train.py

@@ -904,7 +904,7 @@ def train_model(args, output):
 
     if use_cuda:
         model.cuda()
-        if mode_classify == "ot":
+        if cmap_flag and mode_classify == "ot":
             sampler.cuda()
 
     # Train the model