update tests

Author: ZahraGanji

src/evaluation.py

@@ -1,11 +1,13 @@
-import pandas as pd
-from sentence_transformers import SentenceTransformer, util
-import numpy as np
-from rapidfuzz import fuzz
+import random
+
 import Levenshtein
+import numpy as np
 import open_clip
+import pandas as pd
 import torch
-import random
+from rapidfuzz import fuzz
+from sentence_transformers import SentenceTransformer, util
+
 
 def set_seed(seed=42):
     """Set all random seeds for reproducibility"""
@@ -16,47 +18,55 @@ def set_seed(seed=42):
     torch.backends.cudnn.deterministic = True
     torch.backends.cudnn.benchmark = False
 
- 
+
 minilm_model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
-biobert_model = SentenceTransformer("pritamdeka/BioBERT-mnli-snli-scinli-scitail-mednli-stsb")
+biobert_model = SentenceTransformer(
+    "pritamdeka/BioBERT-mnli-snli-scinli-scitail-mednli-stsb"
+)
+
 
 # === Helpers ===
 def preprocess(text):
     return text.lower().strip()
 
+
 def split_and_clean(text):
     if pd.isna(text) or not text.strip():
         return []
     return [preprocess(x) for x in text.split(",") if x.strip()]
 
+
 def post_process(column):
-    return (
-        column.fillna(0.0)
-        .round(4)
-        .astype(float)
-        .replace([np.inf, -np.inf], 0.0)
-    )
+    return column.fillna(0.0).round(4).astype(float).replace([np.inf, -np.inf], 0.0)
+
 
 # === Similarity Functions ===
-       
+
+
 def miniLM_pairwise_avg_similarity(list1, list2):
     if not list1 or not list2:
         return np.nan
     emb1 = minilm_model.encode(list1, convert_to_tensor=True)
     emb2 = minilm_model.encode(list2, convert_to_tensor=True)
     scores = util.cos_sim(emb1, emb2).cpu().numpy()
-    semantic_scores = np.mean(np.max(scores, axis=1))  # average of best matches from list1 to list2
+    semantic_scores = np.mean(
+        np.max(scores, axis=1)
+    )  # average of best matches from list1 to list2
     return semantic_scores
 
+
 def biobert_pairwise_avg_similarity(list1, list2):
     if not list1 or not list2:
         return np.nan
     emb1 = biobert_model.encode(list1, convert_to_tensor=True)
     emb2 = biobert_model.encode(list2, convert_to_tensor=True)
     scores = util.cos_sim(emb1, emb2).cpu().numpy()
-    semantic_scores = np.mean(np.max(scores, axis=1))  # average of best matches from list1 to list2
+    semantic_scores = np.mean(
+        np.max(scores, axis=1)
+    )  # average of best matches from list1 to list2
     return semantic_scores
 
+
 def fuzzy_pairwise_avg_similarity(list1, list2):
     # Fuzzy distance scores
     if not list1 or not list2:
@@ -68,10 +78,16 @@ def edit_pairwise_avg_similarity(list1, list2):
     # edit distance scores
     if not list1 or not list2:
         return np.nan
-    return np.mean([
-        max(1 - (Levenshtein.distance(s1, s2) / max(len(s1), len(s2))) for s2 in list2)
-        for s1 in list1
-    ])
+    return np.mean(
+        [
+            max(
+                1 - (Levenshtein.distance(s1, s2) / max(len(s1), len(s2)))
+                for s2 in list2
+            )
+            for s1 in list1
+        ]
+    )
+
 
 # === Compute Similarity Columns ===
 def compute_similarity_column(sim_func, col_prefix, model_name, reader, df):
@@ -80,47 +96,61 @@ def compute_similarity_column(sim_func, col_prefix, model_name, reader, df):
     df[col_name] = df.apply(
         lambda row: sim_func(
             split_and_clean(row.get(reader, "")),
-            split_and_clean(row.get(f"Protocol {model_name}", ""))
+            split_and_clean(row.get(f"Protocol {model_name}", "")),
         ),
-        axis=1
+        axis=1,
     )
     df[col_name] = post_process(df[col_name])
     return col_name
 
-def evaluation_func(model_name, fine_tuning_method, reader):
+
+def evaluation_func(model_name, fine_tuning_method, layers, reader):
     # === Config ===
     set_seed(42)  # Set seed for reproducibility
     data_path = ""
 
-    if fine_tuning_method == 'none':
+    if fine_tuning_method == "none":
         data_path = "/home/s33zganj/Machine-Learning-Based-Automation-of-MRI-Brain-Protocol-Selection/data/evaluation.xlsx"
-        #if dataset_name == 'evaluation':
+        # if dataset_name == 'evaluation':
         #    data_path = "/home/s33zganj/Machine-Learning-Based-Automation-of-MRI-Brain-Protocol-Selection/data/evaluation.xlsx"
-        #elif dataset_name == 'test':
+        # elif dataset_name == 'test':
         #    data_path = "/home/s33zganj/Machine-Learning-Based-Automation-of-MRI-Brain-Protocol-Selection/data/test.xlsx"
-        
+
     else:
-        data_path = f"/home/s33zganj/Machine-Learning-Based-Automation-of-MRI-Brain-Protocol-Selection/data/evaluation_{fine_tuning_method}_tuning.xlsx"
-        #if dataset_name == 'evaluation':
+        data_path = f"/home/s33zganj/Machine-Learning-Based-Automation-of-MRI-Brain-Protocol-Selection/data/evaluation_{fine_tuning_method}{layers}_tuning.xlsx"
+        # if dataset_name == 'evaluation':
         #    data_path = f"/home/s33zganj/Machine-Learning-Based-Automation-of-MRI-Brain-Protocol-Selection/data/evaluation_{fine_tuning_method}_tuning.xlsx"
-        #elif dataset_name == 'test':
+        # elif dataset_name == 'test':
         #    data_path = f"/home/s33zganj/Machine-Learning-Based-Automation-of-MRI-Brain-Protocol-Selection/data/test_{fine_tuning_method}_tuning.xlsx"
 
-
-
-    #reader = "Protokoll"
+    # reader = "Protokoll"
     # llm = "llama"
 
     # === Load Data ===
     df = pd.read_excel(data_path)
 
-
     # Compute and post-process scores
     cols = []
-    cols.append(compute_similarity_column(miniLM_pairwise_avg_similarity, "MiniLM", model_name, reader, df))
-    cols.append(compute_similarity_column(biobert_pairwise_avg_similarity, "BioBert", model_name, reader, df))
-    cols.append(compute_similarity_column(fuzzy_pairwise_avg_similarity, "fs", model_name, reader, df))
-    cols.append(compute_similarity_column(edit_pairwise_avg_similarity, "ds", model_name, reader, df))
+    cols.append(
+        compute_similarity_column(
+            miniLM_pairwise_avg_similarity, "MiniLM", model_name, reader, df
+        )
+    )
+    cols.append(
+        compute_similarity_column(
+            biobert_pairwise_avg_similarity, "BioBert", model_name, reader, df
+        )
+    )
+    cols.append(
+        compute_similarity_column(
+            fuzzy_pairwise_avg_similarity, "fs", model_name, reader, df
+        )
+    )
+    cols.append(
+        compute_similarity_column(
+            edit_pairwise_avg_similarity, "ds", model_name, reader, df
+        )
+    )
 
     # === Compute Final Scores ===
     print("\n=== Final Average Scores ===")
@@ -131,6 +161,3 @@ def evaluation_func(model_name, fine_tuning_method, reader):
     # === Save Results ===
     df.to_excel(data_path, index=False)
     print(f"\nSaved results to {data_path}")
-
-
-