add first code-by-group model

app.py

@@ -6,7 +6,7 @@ import pickle
 import numpy as np
 import os
 
-from required_classes import BertEmbedder, PredictModel
+from required_classes import *
 
 
 CLS_WEIGHTS = {'mlp': 0.3, 'svc': 0.4, 'xgboost': 0.3}

classifiers/codes_in_groups/L82_code_clf.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:caaa8d9aee3aa4c349740be71f44ae034c36824b6fa9a765fa06dab88af265ea
+size 327292

required_classes.py

@@ -71,4 +71,34 @@ class PredictModel:
         if log:
             print('Start classifier prediction')
         prediction = self.classifier.predict(embeds)
-        return prediction
+        return prediction
+    
+class CustomXGBoost:
+    def __init__(self):
+        self.model = xgb.XGBClassifier()
+        self.classes_ = None
+
+    def fit(self, X, y):
+        self.classes_ = np.unique(y).tolist()
+        y = [self.classes_.index(l) for l in y]
+        self.model.fit(X, y)
+
+    def predict_proba(self, X):
+        pred = self.model.predict_proba(X)
+        return pred
+
+    def predict(self, X):
+        preds = self.model.predict_proba(X)
+        print(np.argmax(preds, axis=1), self.classes_)
+        print(preds.shape, preds[:2])
+        return self.classes_[np.argmax(preds, axis=1)]
+
+class SimpleModel:
+    def __init__(self):
+        self.classes_ = None
+
+    def fit(self, X, y):
+        self.classes_ = [y[0]]
+
+    def predict_proba(self, X):
+        return np.array([1.0])