added older_scripts folder

adithya8 · Apr 24, 2021 · 4e9d262 · 4e9d262
2 parents 5abd42b + ff2ca7c
commit 4e9d262
Show file tree

Hide file tree

Showing 23 changed files with 1,056 additions and 9 deletions.
diff --git a/README.md b/README.md
@@ -1,14 +1,6 @@
 # **Empirical Evaluation of Pre-trained Transformers for Human-Level NLP: The Role of Sample Size and Dimensionality**
 
-## *Adithya V Ganesan, 112683104*
-
-----
-
-The FB dataset detailed in the report was used for this project. The data was dumped in MySQL in order to conduct all the experiments using [DLATK](https://github.com/dlatk/dlatk), a language toolkit. This README file contains the command logs to collect the results for the experiments. All the commands except for the transformer embedding generation was executed by switching to the "dev" branch. The said embedding generation commands were executed checkingout to "dev-transformers" branch.
-
-The messages for domain data was stored in table named D_20 and task data in a table named T_20. The outcomes (age, gen, ext, ope) were stored in a table named 20_outcomes. The database will be referred as "db".
-
----- 
+---
 
 ### **Commands to extract RoBERTa embeddings:**
 
@@ -57,3 +49,14 @@ This command would store the evaluation result for the ten runs in output.txt.
 For classification task the commands have a slight variation. The outcomes fag is changed to appropriate categorical column name. The `--train_reg` and `--predict_reg` are changed to `--train_classifiers` and `--predict_classifiers` respectively. 
 
 ----
+
+| Number of training samples | Demographic Tasks | Personality Tasks | Mental Health Tasks |
+| -------------------------- | :---------------: | :---------------: | :-----------------: |
+| 50                         | 16                | 16                | 16                  |
+| 100			     | 128		 | 16		     | 22		   |
+| 200			     | 512		 | 32		     | 45		   |
+| 500			     | 768		 | 64		     | 64		   |
+| 1000			     | 768		 | 90		     | 64		   |
+
+This work is intended to inform researchers in Computational Social Science a simple way to improve the performance of transformer based models. We find that training PCA on transformer representations using the domain data improves the model performance overall, with evidence of handling longer sequences better than other reduction methods.
+The table above presents a summary of systematic experiments, recommmending the number of dimensions required for given number of samples in each task domain to achieve the best performance.
diff --git a/older_scripts/BERTBDimRedExp.sh b/older_scripts/BERTBDimRedExp.sh
@@ -0,0 +1,129 @@
+#author: @adithya8
+
+if [ "$1" -eq "18" ]; 
+then
+    declare -A db=(["d"]="clp18_adi" ["t"]="tr_a11essays" ["c"]="clp18_id")
+    declare -A dbTables=(["tr_a11"]="tr_a11essays")
+    declare -A lexTables=(["tr_a11"]="tr_a11_bertb_")
+    folderName="clp18_adi"
+    dimRedModel=$2
+    msgk=$3
+    noEval="0"
+
+    if [ "$4" -eq "1" ];
+    then 
+      noEval="1"
+    fi        
+elif [ "$1" -eq "19" ];
+then
+    declare -A db=(["d"]="clp19_adi" ["t"]="task_A" ["c"]="user_id") 
+    declare -A dbTables=(["A"]="task_A" ["C"]="task_Cfil" ["At"]="task_A_title" ["Ct"]="task_Cfil_title")
+    declare -A lexTables=(["A"]="task_A_bert_" ["C"]="task_Cfil_bert_" ["At"]="taskAt_bert_" ["Ct"]="taskCt_bert_")
+    folderName="clp19_adi"
+    dimRedModel=$2
+    msgk=$3
+    titlek=$4         
+
+    noEval="0"
+
+    if [ "$5" -eq "1" ];
+    then 
+      noEval="1"
+    fi    
+else
+    declare -A db=(["d"]="fb20_adi" ["t"]="tr_fb" ["c"]="user_id") 
+    declare -A dbTables=(["tr_fb"]="tr_fb")
+    declare -A lexTables=(["tr_fb"]="tr_fb_bert_")
+    folderName="fb20_adi"
+    dimRedModel=$2
+    msgk=$3    
+    noEval="0"
+
+    if [ "$4" -eq "1" ];
+    then 
+      noEval="1"
+    fi        
+
+fi
+
+
+declare -A dimRedModels=(["fa"]="fa" ["pca"]="pca" ["nmf"]="nmf")
+
+echo "$1_$2_$3"
+echo "${noEval}"
+
+resultFile="BERTb_${dimRedModel}_${msgk}_${titlek}.txt"
+
+echo "$resultFile"
+
+if [ "${noEval}" -eq "0" ];
+then
+    if test -d "~/NLP/ContextualEmbeddingDR/results/${folderName}/BERTb_${dimRedModel}/"; then 
+        echo "Directory Exists"
+    else        
+        eval "mkdir ~/NLP/ContextualEmbeddingDR/results/${folderName}/BERTb_${dimRedModel}/"
+        if test -d "~/NLP/ContextualEmbeddingDR/results/${folderName}/BERTb_${dimRedModel}/"; then 
+            echo "Directory Created"
+        fi
+    fi 
+fi
+
+for i in ${!dbTables[@]}
+ do
+    dbTable=${dbTables[$i]}
+    lexTable=${lexTables[$i]}
+
+    if [[ $dbTable =~ "title" ]]
+    then
+        k=${titlek}
+    else
+        k=${msgk}
+    fi
+
+    lexTableCr="python3 ~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${dbTable} -c ${db["c"]} --group_freq_thresh 0 -f 'feat\$bert_ba_un_memimaL10co\$${dbTable}\$${db["c"]}\$16to16' --fit_reducer --model ${dimRedModel} --n_components ${k} --reducer_to_lexicon ${lexTable}${dimRedModel}${k}"
+    echo "${lexTableCr}"
+
+    if [ "${noEval}" -eq "0" ];
+    then
+        eval "${lexTableCr}"
+    fi
+
+    echo "----------------------------------------------------------------------"
+    weightLex="python3 ~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${dbTable} -c ${db["c"]} --group_freq_thresh 0 --word_table 'feat\$bert_ba_un_memimaL10co\$${dbTable}\$${db["c"]}\$16to16' --add_lex -l ${lexTable}${dimRedModel}${k} --weighted_lex"
+    echo "${weightLex}"
+
+    if [ "${noEval}" -eq "0" ];
+    then 
+        eval "${weightLex}"
+    fi
+
+    echo "----------------------------------------------------------------------"
+done
+if [ "$1" -eq "18" ];
+then
+    finalCommand="python3 ~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${db["t"]} -c ${db["c"]} --group_freq_thresh 0 -f 'feat\$cat_${lexTables["tr_a11"]}${dimRedModel}${msgk}_w\$${db["t"]}\$${db["c"]}\$bert' --outcome_table tr_variables --outcomes a11_bsag_total --combo_test_reg --model ridgehighcv --folds 10 > ~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/BERTb_${dimRedModel}/${resultFile}"
+elif [ "$1" -eq "19" ];
+then 
+    finalCommand="python3 ~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${db["t"]} -c ${db["c"]} --group_freq_thresh 0 -f 'feat\$cat_${lexTables["A"]}${dimRedModel}${msgk}_w\$${db["t"]}\$${db["c"]}\$bert' 'feat\$cat_${lexTables["At"]}${dimRedModel}${titlek}_w\$task_A_title\$user_id\$bert' 'feat\$cat_${lexTables["C"]}${dimRedModel}${msgk}_w\$task_Cfil\$user_id\$bert' 'feat\$cat_${lexTables["Ct"]}${dimRedModel}${titlek}_w\$task_Cfil_title\$user_id\$bert' --outcome_table task_labels_full --outcomes label --nfold_classifiers --model lr  --folds 10 > ~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/BERTb_${dimRedModel}/${resultFile}"
+else
+    finalCommand="python3 ~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${db["t"]} -c ${db["c"]} --group_freq_thresh 0 -f 'feat\$cat_${lexTables["tr_fb"]}${dimRedModel}${msgk}_w\$${db["t"]}\$${db["c"]}\$bert' --outcome_table masterstats_friendratings --outcomes ope con ext agr neu --combo_test_reg --model ridgehighcv --folds 10 > ~/NLP/ContextualEmbeddingDR/results/${folderName}/BERTb_${dimRedModel}/${resultFile}"
+fi
+echo "${finalCommand}" 
+if [ "${noEval}" -eq "0" ];
+then
+    eval "${finalCommand}" 
+fi
+echo "----------------------------------------------------------------------"
+
+if [ "${noEval}" -eq "0" ];
+then
+    if [ "$1" -eq "18" ];
+    then
+        eval "cat ~/NLP/ContextualEmbeddingDR/results/${folderName}/BERTb_${dimRedModel}/${resultFile} | grep \'r\':"
+    elif [ "$1" -eq "19" ];
+    then 
+        eval "cat ~/NLP/ContextualEmbeddingDR/results/${folderName}/BERTb_${dimRedModel}/${resultFile} | grep \'f1\':"
+    else
+        eval "cat ~/NLP/ContextualEmbeddingDR/results/${folderName}/BERTb_${dimRedModel}/${resultFile} | grep \'r\':"
+    fi
+fi
diff --git a/older_scripts/BERTBDimRedTest.sh b/older_scripts/BERTBDimRedTest.sh
@@ -0,0 +1,99 @@
+# author: @adithya8
+
+if [ "$1" -eq "18" ]; 
+then
+    declare -A db=(["d"]="clp18_adi" ["t"]="tr_a11essays" ["c"]="clp18_id" ["o"]="a11_bsag_total")
+    declare -A dbTables=(["te_a11"]="te_a11essays")
+    declare -A lexTables=(["te_a11"]="tr_a11_bertb_")
+    declare -A alpha=(["16"]="100" ["32"]="100" ["64"]="100" ["128"]="1000" ["256"]="1000" ["512"]="10000" ["1024"]="10000" ["2048"]="10000")
+    dimRedModel=$2
+    msgk=$3
+    noEval="0"
+    if [ "$4" -eq "1" ];
+    then
+        noEval="1"
+    fi
+
+    saveModelCommand="python3 ~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${db["t"]} -c ${db["c"]} --group_freq_thresh 0 -f 'feat\$cat_${lexTables["te_a11"]}${dimRedModel}${msgk}_w\$tr_a11essays\$clp18_id\$bert' --outcome_table tr_variables --outcomes a11_bsag_total  --train_regression --model ridge${alpha[${msgk}]} --save_model --picklefile /data/avirinchipur/models/clp${1}_adi/BERTb_${dimRedModel}_${msgk}.pickle"
+else
+    declare -A db=(["d"]="clp19_adi" ["t"]="task_A" ["c"]="user_id" ["o"]="label") 
+    declare -A dbTables=(["Ate"]="task_A_test" ["Cte"]="task_Cfil_test" ["Atte"]="task_A_title_test" ["Ctte"]="task_Cfil_title_test")
+    declare -A lexTables=(["Ate"]="task_A_bert_" ["Cte"]="task_Cfil_bert_" ["Atte"]="taskAt_bert_" ["Ctte"]="taskCt_bert_")
+    dimRedModel=$2
+    msgk=$3
+    titlek=$4         
+    noEval="0"
+
+    if [ "$5" -eq "1" ];
+    then
+        noEval="1"
+    fi
+    saveModelCommand="python3 ~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${db["t"]} -c ${db["c"]} --group_freq_thresh 0 -f 'feat\$cat_${lexTables["Ate"]}${dimRedModel}${msgk}_w\$task_A\$user_id\$bert' 'feat\$cat_${lexTables["Atte"]}${dimRedModel}${titlek}_w\$task_A_title\$user_id\$bert' 'feat\$cat_${lexTables["Cte"]}${dimRedModel}${msgk}_w\$task_Cfil\$user_id\$bert' 'feat\$cat_${lexTables["Ctte"]}${dimRedModel}${titlek}_w\$task_Cfil_title\$user_id\$bert' --outcome_table task_labels_full --outcomes label --train_classifiers --model lr --save_model --picklefile /data/avirinchipur/models/clp${1}_adi/BERTb_${dimRedModel}_${msgk}_${titlek}.pickle"         
+fi
+
+echo "${saveModelCommand}"
+
+if [ "$noEval" -eq "0" ];
+then 
+    eval "${saveModelCommand}"
+fi
+
+
+declare -A dimRedModels=(["fa"]="fa" ["pca"]="pca" ["nmf"]="nmf")
+
+resultFile="BERTb_${dimRedModel}_${msgk}_${titlek}_test.txt"
+
+echo "$resultFile"
+
+echo "$1_$2_$3"
+
+if [ "${noEval}" -eq "0" ];
+then 
+    if test -d "~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/BERTb_${dimRedModel}/"; then 
+        echo "Directory Exists"
+    else        
+        eval "mkdir ~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/BERTb_${dimRedModel}/"
+        if test -d "~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/BERTb_${dimRedModel}/"; then 
+            echo "Directory Created"
+        fi
+    fi 
+fi
+
+for i in ${!dbTables[@]}
+ do
+    dbTable=${dbTables[$i]}
+    lexTable=${lexTables[$i]}
+
+    if [[ $dbTable =~ "title" ]]
+    then
+        k=${titlek}
+    else
+        k=${msgk}
+    fi
+
+    weightLex="python3  ~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${dbTable} -c ${db["c"]} --group_freq_thresh 0 --word_table 'feat\$bert_ba_un_memimaL10co\$${dbTable}\$${db["c"]}\$16to16' --add_lex -l ${lexTable}${dimRedModel}${k} --weighted_lex"
+    echo "${weightLex}"
+
+    if [ "${noEval}" -eq "0" ];
+    then 
+      eval "${weightLex}"
+    fi
+
+    echo "----------------------------------------------------------------------"
+done
+
+if [ "$1" -eq "18" ];
+then
+    finalCommand="python3 ~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${db["t"]} -c ${db["c"]} --group_freq_thresh 0 -f 'feat\$cat_${lexTables["te_a11"]}${dimRedModel}${msgk}_w\$te_a11essays\$${db["c"]}\$bert' --outcome_table te_a11essays_labels --outcomes a11_bsag_total  --predict_regression --model ridge${alpha[${msgk}]} --load --picklefile /data/avirinchipur/models/clp${1}_adi/BERTb_${dimRedModel}_${msgk}.pickle > ~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/BERTb_${dimRedModel}/${resultFile}"
+else 
+    finalCommand="python3 ~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${db["t"]} -c ${db["c"]} --group_freq_thresh 0 -f 'feat\$cat_task_A_bert_${dimRedModel}${msgk}_w\$task_A_test\$user_id\$bert'  'feat\$cat_taskAt_bert_${dimRedModel}${titlek}_w\$task_A_title_test\$user_id\$bert' 'feat\$cat_task_Cfil_bert_${dimRedModel}${msgk}_w\$task_Cfil_test\$user_id\$bert' 'feat\$cat_taskCt_bert_${dimRedModel}${titlek}_w\$task_Cfil_title_test\$user_id\$bert' --outcome_table crowd_test_A_label --outcomes label --predict_classifiers --model lr  --load --picklefile /data/avirinchipur/models/clp${1}_adi/BERTb_${dimRedModel}_${msgk}_${titlek}.pickle > ~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/BERTb_${dimRedModel}/${resultFile}" 
+fi
+echo "${finalCommand}" 
+
+if [ "$noEval" -eq "0" ];
+then 
+    eval "${finalCommand}"
+fi
+
+echo "----------------------------------------------------------------------"
+
diff --git a/older_scripts/ExpIter.sh b/older_scripts/ExpIter.sh
@@ -0,0 +1,44 @@
+#author: @adithya8
+
+declare -a msg=(14 36 64 118 207 386)
+declare -a title=(6 14 26 47 83 154)
+declare -a msgK=(14 36 71 143 286 357 495)
+declare -a titleK=(6 14 29 57 114 143 198)
+declare -a totK=(16 32 64 128 256 512 1024 2048)
+declare -a totK_=( )
+
+
+# Last array index shouldn't apply for XLNet
+
+if [ "$#" -eq "3" ];
+then
+    experiment=$1
+    contextualEmbedding=$2
+    dimRedModel=$3
+elif [ "$#" -eq "2" ]; 
+then 
+    experiment="19"
+    contextualEmbedding=$1
+    dimRedModel=$2
+else
+    echo "Pass Contextual Embedding (BERTB/XLNet), DimRedModel name as arg (pca/fa/nmf/nmfrand) -- Exiting !!!!"
+    exit
+fi
+
+if [ "$1" -ne "19" ]; 
+then
+    msgK=( "${totK[@]}" )
+    titleK=( "${totK_[@]}" )
+fi
+
+arraylength=${#msgK[@]}
+for (( i=1; i<${arraylength}+1; i++ ));
+do
+    echo "bash ~/NLP/ContextualEmbeddingDR/${contextualEmbedding}DimRedExp.sh ${experiment} ${dimRedModel} ${msgK[$i-1]} ${titleK[$i-1]}"
+    echo "----------------------------------------------------------------------"
+    eval "bash ~/NLP/ContextualEmbeddingDR/${contextualEmbedding}DimRedExp.sh ${experiment} ${dimRedModel} ${msgK[$i-1]} ${titleK[$i-1]}"
+done
+
+#echo "f1 scores for increasing k sizes"
+#eval  "cat ./results/XLNet_${dimRedModel}/* | grep \'f1\':"
+#eval "python ~/NLP/ContextualEmbeddingDR/tableMaker.py ${experiment} ${contextualEmbedding} ${dimRedModel}"
diff --git a/older_scripts/XLNetDimRedExp.sh b/older_scripts/XLNetDimRedExp.sh
@@ -0,0 +1,75 @@
+#author: @adithya8
+
+if [ "$1" -eq "18" ]; 
+then
+    declare -A db=(["d"]="clp18_adi" ["t"]="tr_a11essays" ["c"]="clp18_id")
+    declare -A dbTables=(["tr_a11"]="tr_a11essays")
+    declare -A lexTables=(["tr_a11"]="tr_a11_xln_")
+    dimRedModel=$2
+    msgk=$3
+else
+    declare -A db=(["d"]="clp19_adi" ["t"]="task_A" ["c"]="user_id") 
+    declare -A dbTables=(["A"]="task_A" ["C"]="task_Cfil" ["At"]="task_A_title" ["Ct"]="task_Cfil_title")
+    declare -A lexTables=(["A"]="task_A_xln_" ["C"]="task_Cfil_xln_" ["At"]="taskAt_xln_" ["Ct"]="taskCt_xln_")
+    dimRedModel=$2
+    msgk=$3
+    titlek=$4         
+fi
+
+declare -A dimRedModels=(["fa"]="fa" ["pca"]="pca" ["nmf"]="nmf")
+
+resultFile="XLN_${dimRedModel}_${msgk}_${titlek}.txt"
+
+echo "$resultFile"
+
+echo "$1_$2_$3"
+
+if test -d "~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/XLNet_${dimRedModel}/"; then 
+    echo "Directory Exists"
+else        
+    eval "mkdir ~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/XLNet_${dimRedModel}/"
+    if test -d "~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/XLNet_${dimRedModel}/"; then 
+        echo "Directory Created"
+    fi
+fi 
+
+for i in ${!dbTables[@]}
+ do
+    dbTable=${dbTables[$i]}
+    lexTable=${lexTables[$i]}
+
+    if [[ $dbTable =~ "title" ]]
+    then
+        k=${titlek}
+    else
+        k=${msgk}
+    fi
+
+    lexTableCr="~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${dbTable} -c ${db["c"]} --group_freq_thresh 0 -f 'feat\$xlnet_ba_ca_memamiL10co\$${dbTable}\$${db["c"]}\$16to16' --fit_reducer --model ${dimRedModel} --n_components ${k} --reducer_to_lexicon ${lexTable}${dimRedModel}${k}"
+    echo "${lexTableCr}"
+    eval "${lexTableCr}"
+    echo "----------------------------------------------------------------------"
+    weightLex="~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${dbTable} -c ${db["c"]} --group_freq_thresh 0 --word_table 'feat\$xlnet_ba_ca_memamiL10co\$${dbTable}\$${db["c"]}\$16to16' --add_lex -l ${lexTable}${dimRedModel}${k} --weighted_lex"
+    echo "${weightLex}"
+    eval "${weightLex}"
+    echo "----------------------------------------------------------------------"
+done
+
+if [ "$1" -eq "18" ];
+then
+    finalCommand="~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${db["t"]} -c ${db["c"]} --group_freq_thresh 0 -f 'feat\$cat_${lexTables["tr_a11"]}${dimRedModel}${msgk}_w\$${db["t"]}\$${db["c"]}\$xlne' --outcome_table tr_variables --outcomes a11_bsag_total --combo_test_reg --model ridgehighcv --folds 10 > ~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/XLNet_${dimRedModel}/${resultFile}"
+    #saveModelCommand="python3 ~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${db["t"]} -c ${db["c"]} --group_freq_thresh 0 -f 'feat\$cat_${lexTables["tr_a11"]}${dimRedModel}${msgk}_w\$${db["t"]}\$${db["c"]}\$xlne'  --outcome_table task_labels_full --outcomes label --train_classifiers --model ridge --save_model --picklefile ~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/XLNet_${dimRedModel}/xln_${dimRedModel}_${msgk}_${titlek}.pickle"
+else 
+    finalCommand="~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${db["t"]} -c ${db["c"]} --group_freq_thresh 0 -f 'feat\$cat_${lexTables["A"]}${dimRedModel}${msgk}_w\$${dbTables["A"]}\$${db["c"]}\$xlne' 'feat\$cat_${lexTables["At"]}${dimRedModel}${titlek}_w\$${dbTables["At"]}\$${db["c"]}\$xlne' 'feat\$cat_${lexTables["Ct"]}${dimRedModel}${titlek}_w\$${dbTables["Ct"]}\$${db["c"]}\$xlne' 'feat\$cat_${lexTables["C"]}${dimRedModel}${msgk}_w\$${dbTables["C"]}\$${db["c"]}\$xlne' --outcome_table task_labels_full --outcomes label --nfold_classifiers --model lr  --folds 10 > ~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/XLNet_${dimRedModel}/${resultFile}"    
+    #saveModelCommand="python3 ~/dlatk/dlatk/dlatkInterface.py -d ${db["d"]} -t ${db["t"]} -c ${db["c"]} --group_freq_thresh 0 -f 'feat\$cat_task_A_xln_${dimRedModel}${msgk}_w\$task_A\$user_id\$xlne' 'feat\$cat_taskAt_xln_${dimRedModel}${titlek}_w\$task_A_title\$user_id\$xlne' 'feat\$cat_taskCt_xln_${dimRedModel}${titlek}_w\$task_Cfil_title\$user_id\$xlne' 'feat\$cat_task_Cfil_xln_${dimRedModel}${msgk}_w\$task_Cfil\$user_id\$xlne' --outcome_table task_labels_full --outcomes label --train_classifiers --model lr --save_model --picklefile ~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/XLNet_${dimRedModel}/xln_${dimRedModel}_${msgk}_${titlek}.pickle"         
+fi
+echo "${finalCommand}" 
+eval "${finalCommand}" 
+echo "----------------------------------------------------------------------"
+
+if [ "$1" -eq "18" ];
+then
+    eval "cat ~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/XLNet_${dimRedModel}/${resultFile} | grep \'R\':"
+else    
+    eval "cat ~/NLP/ContextualEmbeddingDR/results/clp${1}_adi/XLNet_${dimRedModel}/${resultFile} | grep \'f1\':"
+fi