Update handler.py

handler.py

@@ -8,10 +8,10 @@ import time
 import os
 import torch
 
-def max_pooling(model_output):
+def mean_pooling(model_output):
     # Get dimensions
     Z, Y = len(model_output[0]), len(model_output[0][0])
-  
+
     # Initialize an empty list with length Y (384 in your case)
     output_array = [0.0] * Y 
 
@@ -19,18 +19,15 @@ def max_pooling(model_output):
     for i in range(Z): 
         # Loop over values in innermost arrays (Y)
         for j in range(Y): 
-            # If value is greater than current max, update max
-            if model_output[0][i][j] > output_array[j]: 
-                output_array[j] = model_output[0][i][j]
+            # Accumulate values
+            output_array[j] += model_output[0][i][j]
+
+    # Compute mean
+    output_array = [val / Z for val in output_array]
     
     return output_array
     
-#Mean Pooling - Take attention mask into account for correct averaging
-def mean_pooling(model_output, attention_mask):
-    token_embeddings = model_output[0] #First element of model_output contains all token embeddings
-    input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
-    return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)
-    
+
 class EndpointHandler():
     def __init__(self, path=""):
         print("HELLO THIS IS THE CWD:", os.getcwd())
@@ -73,6 +70,6 @@ class EndpointHandler():
             # embedding = mean_pooling(model_output, encoded_input['attention_mask'])
             print("F")
             
-            sentence_embeddings.append(max_pooling(model_output))
+            sentence_embeddings.append(mean_pooling(model_output))
             print("G")
         return sentence_embeddings