aimodelshare 0.1.12__py3-none-any.whl → 0.1.64__py3-none-any.whl

aimodelshare/main/eval_lambda.txt

@@ -20,6 +20,45 @@ logger = logging.getLogger(__name__)
 # from s3connect import get_ytestdata, get_onnx_mem, get_onnx_temp
 
 
+####################################################################
+##################### Helper Functions #############################
+
+def _ensure_df(obj):
+    """
+    Safely convert various input types to pandas DataFrame.
+    Handles dict, list, and DataFrame inputs with robust error handling.
+    
+    For dict inputs, always treats them as a single row to maintain consistent
+    behavior and avoid unexpected flattening of nested structures.
+    
+    Args:
+        obj: Input object (DataFrame, dict, list, or other)
+    
+    Returns:
+        pandas.DataFrame
+    """
+    import pandas as pd
+    
+    if isinstance(obj, pd.DataFrame):
+        return obj
+    
+    if isinstance(obj, dict):
+        # Always treat dict as single row to maintain consistent structure
+        return pd.DataFrame([obj])
+    
+    if isinstance(obj, list):
+        if len(obj) == 0:
+            return pd.DataFrame()
+        # If all elements are dicts, create DataFrame from list of dicts
+        if all(isinstance(x, dict) for x in obj):
+            return pd.DataFrame(obj)
+        # Otherwise, create single-column DataFrame
+        return pd.DataFrame({'value': obj})
+    
+    # For any other type, return empty DataFrame
+    return pd.DataFrame()
+
+
 ####################################################################
 ########################### main handler ###########################
 
@@ -63,7 +102,7 @@ def handler(event, context):
         return exdata_dict 
 
     idtoken=event['requestContext']['authorizer']['principalId']
-    decoded = jwt.decode(idtoken, options={"verify_signature": False})  # works in PyJWT < v2.0
+    decoded = jwt.decode(idtoken,options={"verify_signature": False,"verify_aud": False}) # works in PyJWT < v2.0
     email=decoded['email']
     print(email)
     submission_type = body.get("submission_type", None)
@@ -859,6 +898,7 @@ def model_from_string(model_type):
         'BayesianRidge': 'sklearn.linear_model',
         'BernoulliNB': 'sklearn.naive_bayes',
         'BernoulliRBM': 'sklearn.neural_network',
+        'CalibratedClassifierCV': 'sklearn.calibration',
         'CategoricalNB': 'sklearn.naive_bayes',
         'ClassifierMixin': 'sklearn.naive_bayes',
         'ComplementNB': 'sklearn.naive_bayes',
@@ -896,6 +936,7 @@ def model_from_string(model_type):
         'LassoLars': 'sklearn.linear_model',
         'LassoLarsCV': 'sklearn.linear_model',
         'LassoLarsIC': 'sklearn.linear_model',
+        'LinearDiscriminantAnalysis': 'sklearn.discriminant_analysis',
         'LinearRegression': 'sklearn.linear_model',
         'LinearSVC': 'sklearn.svm',
         'LinearSVR': 'sklearn.svm',
@@ -930,6 +971,7 @@ def model_from_string(model_type):
         'PassiveAggressiveRegressor': 'sklearn.linear_model',
         'Perceptron': 'sklearn.linear_model',
         'PoissonRegressor': 'sklearn.linear_model',
+        'QuadraticDiscriminantAnalysis': 'sklearn.discriminant_analysis',
         'RANSACRegressor': 'sklearn.linear_model',
         'RadiusNeighborsClassifier': 'sklearn.neighbors',
         'RadiusNeighborsRegressor': 'sklearn.neighbors',
@@ -984,13 +1026,25 @@ def get_leaderboard(task_type="classification", verbose=3, columns=None, private
 
     s3 = boto3.resource("s3")
     bucketres = s3.Bucket("$bucket_name")
-    with open("/tmp/"+mastertable_path.split("/")[-1]+".csv", "wb") as lbfo:
-      bucketres.download_fileobj("$unique_model_id/"+mastertable_path+".csv", lbfo)              
-      
-      
-      
-    leaderboard = pd.read_csv("/tmp/"+mastertable_path.split("/")[-1]+".csv", sep="\t")
-    currentversions=leaderboard['version']
+    
+    # Defensive read of master leaderboard file
+    try:
+        with open("/tmp/"+mastertable_path.split("/")[-1]+".csv", "wb") as lbfo:
+            bucketres.download_fileobj("$unique_model_id/"+mastertable_path+".csv", lbfo)
+        leaderboard = pd.read_csv("/tmp/"+mastertable_path.split("/")[-1]+".csv", sep="\t")
+    except Exception as e:
+        print(f"Warning: Could not read master leaderboard file: {e}")
+        leaderboard = pd.DataFrame()
+    
+    # Ensure leaderboard is a DataFrame
+    leaderboard = _ensure_df(leaderboard)
+    
+    # Get current versions if available
+    try:
+        currentversions=leaderboard['version']
+    except KeyError:
+        currentversions=[]
+    
     print("current versions:")
     print(list(currentversions))
     allversions = [sub.split('_v')[1].split('.')[0] for sub in newleaderboarddata]
@@ -1002,12 +1056,26 @@ def get_leaderboard(task_type="classification", verbose=3, columns=None, private
     #TODO: check if items in leaderboard, if so, then do following
     if len(missingincurrent_leaderboard)>0:
         for i in missingincurrent_leaderboard:
-            with open("/tmp/"+mastertable_path.split("/")[-1]+"_v"+str(i)+".csv", "wb") as lbfo:
-              bucketres.download_fileobj("$unique_model_id/"+mastertable_path+"_v"+str(i)+".csv", lbfo)
-            newleaderboard = pd.read_csv("/tmp/"+mastertable_path.split("/")[-1]+"_v"+str(i)+".csv", sep="\t")
-            newleaderboard.drop(newleaderboard.filter(regex="Unname"),axis=1, inplace=True)
+            try:
+                with open("/tmp/"+mastertable_path.split("/")[-1]+"_v"+str(i)+".csv", "wb") as lbfo:
+                    bucketres.download_fileobj("$unique_model_id/"+mastertable_path+"_v"+str(i)+".csv", lbfo)
+                newleaderboard = pd.read_csv("/tmp/"+mastertable_path.split("/")[-1]+"_v"+str(i)+".csv", sep="\t")
+                newleaderboard.drop(newleaderboard.filter(regex="Unname"),axis=1, inplace=True)
+            except Exception as e:
+                print(f"Warning: Could not read leaderboard for version {i}: {e}")
+                continue
             
-            leaderboard=leaderboard.append(newleaderboard).drop_duplicates()
+            # Ensure newleaderboard is a DataFrame before concat
+            newleaderboard = _ensure_df(newleaderboard)
+            
+            # Use pd.concat instead of deprecated append
+            leaderboard = pd.concat([leaderboard, newleaderboard], ignore_index=True)
+            
+        # Deduplicate once after all concatenations, keeping last occurrence
+        if 'version' in leaderboard.columns:
+            leaderboard = leaderboard.drop_duplicates(subset=['version'], keep='last')
+        else:
+            leaderboard = leaderboard.drop_duplicates(keep='last')
             
         leaderboard.drop(leaderboard.filter(regex="Unname"),axis=1, inplace=True)
     #save new leaderboard here