Fix cache hit rate when chunked prefill (#2555)

python/sglang/srt/managers/schedule_policy.py

@@ -248,7 +248,7 @@ class PrefillAdder:
         self.can_run_list.append(req)
 
         self._prefill_one_req(
-            len(req.prefix_indices),
+            0,
             req.extend_input_len,
             (
                 min(req.sampling_params.max_new_tokens, CLIP_MAX_NEW_TOKENS_ESTIMATION)

python/sglang/srt/managers/scheduler.py

@@ -629,7 +629,6 @@ class Scheduler:
         self.waiting_queue.append(req)
 
     def log_prefill_stats(self, adder, can_run_list, running_bs, has_being_chunked):
-        if isinstance(self.tree_cache, RadixCache):
         self.tree_cache_metrics["total"] += (
             adder.log_input_tokens + adder.log_hit_tokens
         ) / 10**9
@@ -637,8 +636,6 @@ class Scheduler:
         tree_cache_hit_rate = (
             self.tree_cache_metrics["hit"] / self.tree_cache_metrics["total"]
         )
-        else:
-            tree_cache_hit_rate = 0.0
 
         num_used = self.max_total_num_tokens - (
             self.token_to_kv_pool.available_size() + self.tree_cache.evictable_size()