⚡ Bolt: Replace O(N^2) list pop(0) with slice extend in entropy calculation

.jules/bolt.md

@@ -0,0 +1,3 @@
+## 2025-05-05 - Replacing O(N^2) list pop(0) in fastdeploy/model_executor/entropy_utils.py
+**Learning:** Found O(N^2) list operations where `entropy.pop(0)` was called inside nested loops over batch size and sequence lengths.
+**Action:** Replace `pop(0)` within a loop with an index tracker and bulk addition via `.extend(list[start:end])`.

fastdeploy/model_executor/entropy_utils.py

@@ -49,9 +49,13 @@ def calculate_logits_entropy(logits, share_inputs, temperature):
     entropy_tensor = get_entropy(logits)
     entropy = entropy_tensor.tolist()
 
+    entropy_idx = 0
     for i in range(real_bsz):
-        for _ in range(real_seq_lens[i]):
-            share_inputs["entropy_list"][i].append(entropy.pop(0))
+        seq_len = int(real_seq_lens[i])
+        if seq_len > 0:
+            share_inputs["entropy_list"][i].extend(entropy[entropy_idx : entropy_idx + seq_len])
+            entropy_idx += seq_len
+
         if (
             share_inputs["stop_flags"][i]
             and share_inputs["seq_lens_decoder"][i] != 0
@@ -92,9 +96,13 @@ def speculate_calculate_logits_entropy(logits, share_inputs, temperature):
     entropy_tensor = get_entropy(accepted_logits)
     entropy = entropy_tensor.tolist()
 
+    entropy_idx = 0
     for i in range(real_bsz):
-        for _ in range(share_inputs["accept_num"][i]):
-            share_inputs["entropy_list"][i].append(entropy.pop(0))
+        accept_num = int(share_inputs["accept_num"][i])
+        if accept_num > 0:
+            share_inputs["entropy_list"][i].extend(entropy[entropy_idx : entropy_idx + accept_num])
+            entropy_idx += accept_num
+
         if (
             share_inputs["stop_flags"][i]
             and share_inputs["seq_lens_decoder"][i] != 0