[Feature]CP support data clear (#4214)

* Update serving_chat.py

* Update serving_completion.py

* Update serving_completion.py

* mv connection_manager init

* [BugFix] fix kv cache

* fix format

* [Feature] support clear data

---------

Co-authored-by: Yuanle Liu <yuanlehome@163.com>
Co-authored-by: RAM <gstian5555@outlook.com>

fastdeploy/output/token_processor.py

@@ -464,6 +464,31 @@ class TokenProcessor:
         main_process_metrics.request_inference_time.observe(current_time - task.inference_start_time)
         main_process_metrics.request_generation_tokens.observe(self.tokens_counter[task.request_id])
 
+    def clear_data(self):
+        if envs.ENABLE_V1_KVCACHE_SCHEDULER:
+            self.resource_manager.clear_data()
+        for i in range(self.cfg.max_num_seqs):
+            if self.resource_manager.stop_flags[i]:
+                continue
+            task = self.resource_manager.tasks_list[i]
+            result = RequestOutput(
+                request_id=task.request_id,
+                outputs=CompletionOutput(
+                    index=i,
+                    send_idx=self.tokens_counter[task.request_id],
+                    token_ids=task.eos_token_ids,
+                    draft_token_ids=[],
+                ),
+                finished=True,
+                metrics=RequestMetrics(
+                    arrival_time=time.time(),
+                    request_start_time=task.arrival_time,
+                ),
+            )
+            is_prefill = task.disaggregate_info is not None and task.disaggregate_info["role"] == "prefill"
+            self._recycle_resources(task.request_id, i, task, result, is_prefill)
+            llm_logger.warning(f"clear data for task {task.request_id}")
+
     def _record_speculative_decoding_mertics(self, accept_num):
         """Record metrics of speculative decoding"""
         if not hasattr(main_process_metrics, "spec_decode_draft_acceptance_rate"):