sync commit

Author: jcapp4

chebai/loss/boost_bce.py

@@ -0,0 +1,32 @@
+import torch
+import sys
+sys.path.insert(1,'/home/programmer/Bachelorarbeit/python-chebai')
+
+import extras.weight_loader as f
+
+
+class BCE_Point_boosting(torch.nn.BCEWithLogitsLoss):
+    
+    def __init__(
+        self,
+        **kwargs
+    ):
+        super().__init__(reduction=None,**kwargs)
+
+    def forward(
+            self,
+            input: torch.Tensor,
+            target: torch.Tensor,
+            **kwargs
+            
+    )-> torch.Tensor:
+        weights = kwargs['weights']
+        loss = super().forward(input=input,target=target)
+        weights_tensor = f.create_weight_tensor(weights)
+        loss_scaled = torch.matmul(weights_tensor,loss)
+        return torch.mean(loss_scaled)
+
+        
+        
+        
+        

extras/weight_loader.py

@@ -60,7 +60,8 @@ def find_weight(path:str,ident:int)-> float:
 
     print(f"{ident} is not in file ")
 
-
+#to do 
+# return should be a tuple of weigths matching the sequenece of the target and label tensor
 def create_data_weights(batchsize:int,dim:int,weights:dict[str,float],idents:tuple[int,...])-> torch.tensor:
     weight = torch.empty(batchsize,dim)
     index = 0
@@ -76,6 +77,9 @@ def create_data_weights(batchsize:int,dim:int,weights:dict[str,float],idents:tup
 def testing():
     print("hello world")
 
+#create a tensor that is size (1,n) where n is the amout of classes being predicted
+def create_weight_tensor(weight:float)-> torch.tensor:
+    pass