[Optimize] optimize prefix cache in release22 (#3889)

* optimize prefix cache in release22

* optimize prefix cache in release22

* fix worker

* fix

* fix

---------

Co-authored-by: Jiang-Jia-Jun <[email protected]>

Author: rainyfly

fastdeploy/cache_manager/prefix_cache_manager.py

@@ -257,7 +257,12 @@ def can_allocate_gpu_blocks(self, num_blocks: int):
         Check if num_blocks gpu blocks can be allocated.
         """
         if len(self.gpu_free_block_list) < num_blocks:
-            return False
+            if self.cache_config.enable_prefix_caching:
+                self.free_block_ids(num_blocks)
+            if len(self.gpu_free_block_list) < num_blocks:
+                return False
+            else:
+                return True
         else:
             return True
 
@@ -448,7 +453,7 @@ def get_required_block_num(self, input_token_num, block_size):
         """
         return (input_token_num + block_size - 1) // block_size
 
-    def update_cache_blocks(self, task, block_size):
+    def update_cache_blocks(self, task, block_size, num_computed_tokens):
         """
         update cache blocks for a task.
         # TODO(chengyanfu): support async update
@@ -459,12 +464,15 @@ def update_cache_blocks(self, task, block_size):
         """
         try:
             req_id = task.request_id
-            num_cached_tokens = task.num_cached_tokens
             block_tables = task.block_tables
 
-            last_node, input_ids = self.cache_info[req_id]
-            left_input_ids = input_ids[num_cached_tokens:]
+            last_node, num_cached_tokens = self.cache_info[req_id]
+            input_ids = task.prompt_token_ids + task.output_token_ids
+            can_cache_computed_tokens = num_computed_tokens - num_computed_tokens % block_size
+            left_input_ids = input_ids[num_cached_tokens:can_cache_computed_tokens]
             gpu_extra_block_ids = block_tables[num_cached_tokens // block_size :]
+            if req_id in self.leaf_req_map[last_node]:  # delete old leaf record, update later
+                self.leaf_req_map[last_node].remove(req_id)
 
             with self.request_release_lock:
                 current_time = time.time()
@@ -480,7 +488,8 @@ def update_cache_blocks(self, task, block_size):
                 )
                 self.req_leaf_map[req_id] = leaf_node
                 self.leaf_req_map[leaf_node].add(req_id)
-                self.cache_info[req_id] = (leaf_node, input_ids)
+                self.cache_info[req_id] = (leaf_node, can_cache_computed_tokens)
+                task.cached_block_num = can_cache_computed_tokens // block_size
         except Exception as e:
             logger.error(f"update_cache_blocks, error: {type(e)} {e}, {str(traceback.format_exc())}")
             raise e
@@ -508,7 +517,7 @@ def request_match_blocks(self, task, block_size, *args):
                 hit_info["gpu_cache_blocks"] = 0
                 hit_info["cpu_cache_blocks"] = 0
                 self.metrics.req_count += 1
-                input_ids = task.prompt_token_ids
+                input_ids = task.prompt_token_ids + task.output_token_ids
                 req_id = task.request_id
                 logger.info(f"request_match_blocks: start to allocate blocks for req_id {req_id}")
                 input_token_num = len(input_ids)
@@ -546,9 +555,6 @@ def request_match_blocks(self, task, block_size, *args):
                         "request_match_blocks: Not enough GPU memory to allocate cache for matched CPU Cache"
                     )
 
-                #  record request cache info
-                self.cache_info[req_id] = (match_block_node, input_ids)
-
                 # 3. update metrics
                 matched_token_num = gpu_match_token_num + cpu_match_token_num
                 common_block_ids = match_gpu_block_ids + gpu_recv_block_ids
@@ -571,6 +577,9 @@ def request_match_blocks(self, task, block_size, *args):
                 # set leaf node temporarily, then update it in update_cache_blocks
                 self.req_leaf_map[req_id] = match_block_node
                 self.leaf_req_map[match_block_node].add(req_id)
+                #  record request cache info
+                self.cache_info[req_id] = (match_block_node, matched_token_num)
+                task.cached_block_num = matched_token_num // block_size
                 return common_block_ids, matched_token_num, hit_info
             except Exception as e:
                 logger.error(f"request_match_blocks: request_block_ids: error: {type(e)} {e}")
@@ -687,6 +696,11 @@ def release_block_ids_async(self, task):
         """
         return self.executor_pool.submit(self.release_block_ids, task)
 
+    def free_block_ids(self, need_block_num):
+        self.free_block_ids_async(need_block_num)
+        while (self.gpu_free_task_future is not None) and (not self.gpu_free_task_future.done()):
+            time.sleep(0.001)
+
     def release_block_ids(self, task):
         """
         release block ids
@@ -1108,15 +1122,6 @@ def _update_matched_node_info(self, req_id, last_node, current_time):
             node.req_id_set.add(req_id)
             node = node.parent
 
-    def decrease_request_share_count(self, req_id):
-        """
-        Decrease node shared count
-        """
-        node, input_ids = self.cache_info[req_id]
-        while node != self.radix_tree_root:
-            node.decrement_shared_count()
-            node = node.parent
-
     def build_path(
         self,
         req_id,

fastdeploy/engine/common_engine.py

@@ -527,9 +527,8 @@ def _fetch_request():
                 self.cfg.max_prefill_batch,
             )
 
-            self.resource_manager.check_and_free_block_tables()
             tasks = self.scheduler.get_requests(
-                available_blocks=self.resource_manager.available_block_num(),
+                available_blocks=self.cfg.cache_config.max_block_num_per_seq,
                 block_size=self.cfg.cache_config.block_size,
                 reserved_output_blocks=self.cfg.cache_config.enc_dec_block_num,
                 max_num_batched_tokens=self.cfg.max_model_len,

fastdeploy/engine/sched/resource_manager_v1.py

@@ -84,7 +84,6 @@ def allocated_slots(self, request: Request):
         return len(request.block_tables) * self.config.cache_config.block_size
 
     def get_new_block_nums(self, request: Request, num_new_tokens: int):
-        self.check_and_free_block_tables()
         return (
             request.num_computed_tokens + num_new_tokens + self.config.cache_config.block_size - 1
         ) // self.config.cache_config.block_size - len(request.block_tables)
@@ -119,7 +118,7 @@ def _trigger_preempt(self, request, num_new_blocks, preempted_reqs, scheduled_re
                 preempted_req.status = RequestStatus.PREEMPTED
                 preempted_req.num_computed_tokens = 0
                 self._free_blocks(preempted_req)
-                preempted_req.prefill_block_num = None
+                preempted_req.cached_block_num = 0
                 self.to_be_rescheduled_request_id_set.add(preempted_req.request_id)
                 preempted_reqs.append(preempted_req)
                 scheduled_reqs.append(self._prepare_preempt_task(preempted_req))
@@ -282,14 +281,6 @@ def schedule(self):
                 if request.num_computed_tokens >= request.need_prefill_tokens:  # to be decoding
                     if request.num_total_tokens > request.need_prefill_tokens:  # has generated tokens
                         request.num_computed_tokens = request.num_total_tokens - 1
-                    else:  # prefill finished
-                        if (
-                            self.config.cache_config.enable_prefix_caching
-                            and request.get("prefill_block_num", None) is None
-                        ):
-                            # update prefill cache blocks for prefix caching
-                            request.prefill_block_num = len(request.block_tables)
-                            self.cache_manager.update_cache_blocks(request, self.config.cache_config.block_size)
                     if (
                         self.allocated_slots(request) - request.num_total_tokens
                         <= self.config.cache_config.prealloc_dec_block_slot_num_threshold
@@ -339,6 +330,10 @@ def schedule(self):
                         scheduled_reqs.append(self._prepare_prefill_task(request, num_new_tokens))
                     token_budget -= num_new_tokens
                     request.num_computed_tokens += num_new_tokens
+                    if self.config.cache_config.enable_prefix_caching:
+                        self.cache_manager.update_cache_blocks(
+                            request, self.config.cache_config.block_size, request.num_computed_tokens
+                        )
                 req_index += 1
             # schedule the WAITING requests.
             if not preempted_reqs:
@@ -371,6 +366,10 @@ def schedule(self):
                             request.schedule_start_time = time.time()
                             token_budget -= num_new_tokens
                             request.num_computed_tokens += num_new_tokens
+                            if self.config.cache_config.enable_prefix_caching:
+                                self.cache_manager.update_cache_blocks(
+                                    request, self.config.cache_config.block_size, request.num_computed_tokens
+                                )
                             request.status = RequestStatus.RUNNING
                             main_process_metrics.num_requests_waiting.dec(1)
                             main_process_metrics.num_requests_running.inc(1)
@@ -403,6 +402,10 @@ def schedule(self):
                             scheduled_reqs.append(self._prepare_prefill_task(request, num_new_tokens))
                             token_budget -= num_new_tokens
                             request.num_computed_tokens += num_new_tokens
+                            if self.config.cache_config.enable_prefix_caching:
+                                self.cache_manager.update_cache_blocks(
+                                    request, self.config.cache_config.block_size, request.num_computed_tokens
+                                )
                             request.status = RequestStatus.RUNNING
                             main_process_metrics.num_requests_waiting.dec(1)
                             main_process_metrics.num_requests_running.inc(1)
@@ -447,7 +450,7 @@ def get_prefix_cached_blocks(self, request: Request):
 
             matched_block_num = len(common_block_ids)
             no_cache_block_num = self.cache_manager.get_required_block_num(
-                request.prompt_token_ids_len - matched_token_num,
+                request.need_prefill_tokens - matched_token_num,
                 self.config.cache_config.block_size,
             )
 
@@ -463,7 +466,7 @@ def get_prefix_cached_blocks(self, request: Request):
             main_process_metrics.prefix_gpu_cache_token_num.inc(request.gpu_cache_token_num)
             main_process_metrics.prefix_cpu_cache_token_num.inc(request.cpu_cache_token_num)
 
-            if matched_token_num == request.prompt_token_ids_len:
+            if matched_token_num == request.need_prefill_tokens:
                 request.num_computed_tokens = matched_token_num - self.config.cache_config.block_size
                 request.skip_allocate = True
             else:
@@ -481,16 +484,8 @@ def add_request(self, request: Request) -> None:
 
     def _free_blocks(self, request: Request):
         if self.config.cache_config.enable_prefix_caching:
-            # TODO(chengyanfu): support cache ouput blocks for prefix caching
-            if request.get("prefill_block_num", None) is None:
-                leaf_node = self.cache_manager.req_leaf_map[request.request_id]
-                self.cache_manager.decrease_request_share_count(request.request_id)
-                self.cache_manager.free_nodes_directly(leaf_node)
-                self.cache_manager.recycle_gpu_blocks(request.block_tables[request.cache_info[0] :])
-
-            else:
-                self.cache_manager.release_block_ids_async(request)
-                self.cache_manager.recycle_gpu_blocks(request.block_tables[request.prefill_block_num :])
+            self.cache_manager.release_block_ids(request)
+            self.cache_manager.recycle_gpu_blocks(request.block_tables[request.cached_block_num :])
         else:
             self.cache_manager.recycle_gpu_blocks(request.block_tables)
         request.block_tables = []

fastdeploy/worker/gpu_model_runner.py

@@ -1347,6 +1347,7 @@ def _get_skip_idx(self, model_forward_batch: Optional[List[Request]] = None):
         if (
             not self.cache_config.enable_chunked_prefill
             or self.guided_backend is None
+            or model_forward_batch is None
             or envs.ENABLE_V1_KVCACHE_SCHEDULER
         ):
             return skip_idx_list
@@ -1549,7 +1550,7 @@ def _add_cache(self, model_forward_batch) -> None:
         """
         Add cache for guided decoding.
         """
-        if self.guided_backend is None:
+        if self.guided_backend is None or model_forward_batch is None:
             return
 
         for request in model_forward_batch: