ROC curve for each class

Question

Code:

df = pd.read_csv(r"model_data_TBI_3(o).csv", index_col= ["X","Y"])
X = df.drop("class", axis= "columns")
y = df["class"]
array  = y.values
y = array[: :]
  
y = label_binarize(y, classes= [1,2,3], pos_label= 1, neg_label= 0)
n_classes = y.shape[1]
print(n_classes)
X_train, X_test, y_train, y_test = train_test_split(X, y,random_state=45, train_size=0.10)
mm = MinMaxScaler()
mm.fit(X_train)
X_train_mm = mm.transform(X_train)
X_test_mm = mm.transform(X_test)
svm_tuned = SVC(kernel= "rbf", C=2.33, gamma= 0.00046, probability= True,
                break_ties= False, decision_function_shape="ovo",
                random_state=1, shrinking=True, tol=0.12,
                class_weight={1:100, 1:75, 2:55})
y_score = svm_tuned.fit(X_train_mm, y_train).decision_function(X_test_mm)

Result:

ValueError: y should be a 1d array, got an array of shape (329, 3) instead.

I want to plot the ROC curve of my svm model. My data has 1,2,3 classes. I binarized them using label binarizer but still getting this error. The traceback error is at y_score. My request to those who wanna clear my doubt or solution please dont send me the code of iris data i can look it up on sklearn website just give me explanation or write your code as a solution to this problem. I am really thankful to those who will answer. Sorry if i make any mistake in posting questions i am new to stackoverflow and new to python & machine learning. Thank you