files

pert.py

@@ -45,7 +45,9 @@ perturbation_descriptions = {
     "Casing Class Variable Modification": "Generates lexical variations of class and variable names with different casing.",
     "Onecase Modification": "Generates lexical variations of class and variable names with just 1 letter uppercase wither for class anme or variable name.",
     "Example Modification": "Generates lexical variations of class and variable names with Example being the class name and example being the variable name or vice versa.",
-    "Finetuned on compile error": "Clusters based on finetuned codebert model on compile errors"
+    "Finetuned on compile error": "Clusters based on finetuned codebert model on compile errors",
+    "Finetuned on language classification": "Clusters based on finetuned codebert model on language classification",
+
 }
 
 def compute_and_log_csi(file_orig, file_pert, perturbation_name, output_csv="results/csi_summary.csv"):
@@ -170,4 +172,11 @@ compute_and_log_csi(
     perturbation_name="Finetuned on compile error",
     output_csv="results/csi_summary.csv"
 )
+
+compute_and_log_csi(
+    "codenet_4000_finetuned_language_classification/java/layer12/kmeans/clusters-kmeans-350.txt",
+    "codenet_4000_finetuned_language_classification/java/layer12/kmeans/clusters-kmeans-350.txt",
+    perturbation_name="Finetuned on language classification",
+    output_csv="results/csi_summary.csv"
+)
 # You can now call compute_and_log_csi again and again for other perturbations!

results/csi_summary.csv

@@ -12,3 +12,4 @@ Casing Class Variable Modification,0.6722713965133429,0.3277286034866571,Generat
 Onecase Modification,0.665697304921991,0.334302695078009,Generates lexical variations of class and variable names with just 1 letter uppercase wither for class anme or variable name.
 Example Modification,1.0,0.0,Generates lexical variations of class and variable names with Example being the class name and example being the variable name or vice versa.
 Finetuned on compile error,1.0,0.0,Clusters based on finetuned codebert model on compile errors
+Finetuned on language classification,1.0,0.0,Clusters based on finetuned codebert model on language classification