[fix] fix ep group all-reduce (#4140)

* [fix] fix ep group all-reduce

* [fix] fix clear/update lock not working when workers > 1

* [chore] add preemption triggered info log

* [fix] fix code style

* fix model_weights_signal (#4092)

* fix model_weights_signal

---------

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

fastdeploy/config.py

@@ -352,8 +352,12 @@ class ParallelConfig:
         )
         dist.collective._set_custom_gid(None)
         # same ep group id
+        # dist.collective._set_custom_gid(self.data_parallel_size + tp_gid_offset)
+        # self.ep_group = dist.new_group(range(self.expert_parallel_size))
+        if self.enable_expert_parallel:
             dist.collective._set_custom_gid(self.data_parallel_size + tp_gid_offset)
             self.ep_group = dist.new_group(range(self.expert_parallel_size))
+            dist.collective._set_custom_gid(None)
         logger.info(
             f"data_parallel_size: {self.data_parallel_size}, tensor_parallel_size: {self.tensor_parallel_size}, expert_parallel_size: {self.expert_parallel_size}, data_parallel_rank: {self.data_parallel_rank}, tensor_parallel_rank: {self.tensor_parallel_rank}, expert_parallel_rank: {self.expert_parallel_rank}, tp_group: {self.tp_group}."
         )

fastdeploy/engine/sched/resource_manager_v1.py

@@ -120,6 +120,7 @@ class ResourceManagerV1(ResourceManager):
                 self._free_blocks(preempted_req)
                 preempted_req.cached_block_num = 0
                 self.to_be_rescheduled_request_id_set.add(preempted_req.request_id)
+                llm_logger.info(f"Preemption is triggered! Preempted request id: {preempted_req.request_id}")
                 preempted_reqs.append(preempted_req)
                 scheduled_reqs.append(self._prepare_preempt_task(preempted_req))
                 main_process_metrics.num_requests_waiting.inc(1)

fastdeploy/entrypoints/engine_client.py

@@ -16,12 +16,12 @@
 
 import inspect
 import os
-import threading
 import time
 import traceback
 import uuid
 
 import numpy as np
+from filelock import FileLock
 
 from fastdeploy import envs
 from fastdeploy.config import ModelConfig
@@ -132,7 +132,7 @@ class EngineClient:
             pid, max_connections=int(os.getenv("FD_DEALER_CONNECTIONS", 50))
         )
         self.connection_initialized = False
-        self.clear_update_lock = threading.Lock()
+        self.clear_update_lock = FileLock(f"/tmp/fd_weight_clear_update_lock__pid{pid}_port{port}.lock")
 
     def create_zmq_client(self, model, mode):
         """
@@ -351,7 +351,9 @@ class EngineClient:
             if self.model_weights_status_signal.value[0] == ModelWeightsStatus.NORMAL:
                 return True, ""
             if self.model_weights_status_signal.value[0] == ModelWeightsStatus.UPDATING:
-                return False, "updating model weight already"
+                return False, "worker is updating model weight already"
+            if self.model_weights_status_signal.value[0] == ModelWeightsStatus.CLEARING:
+                return False, "worker is clearing model weight, cannot update now"
 
             self.model_weights_status_signal.value[0] = ModelWeightsStatus.UPDATING
             if self.enable_prefix_caching or self.enable_splitwise:
@@ -395,7 +397,9 @@ class EngineClient:
             if self.model_weights_status_signal.value[0] == ModelWeightsStatus.CLEARED:
                 return True, ""
             if self.model_weights_status_signal.value[0] == ModelWeightsStatus.CLEARING:
-                return False, "clearing model weight already"
+                return False, "worker is clearing model weight already"
+            if self.model_weights_status_signal.value[0] == ModelWeightsStatus.UPDATING:
+                return False, "worker is updating model weight, cannot clear now"
 
             self.model_weights_status_signal.value[0] = ModelWeightsStatus.CLEARING
             if self.enable_prefix_caching or self.enable_splitwise:

fastdeploy/model_executor/layers/moe/fused_moe_cutlass_backend.py

@@ -297,7 +297,7 @@ class CutlassMoEMethod(UnquantizedFusedMoEMethod):
         )
 
         if layer.reduce_results and layer.tp_size > 1:
-            tensor_model_parallel_all_reduce(fused_moe_out)
+            tensor_model_parallel_all_reduce(fused_moe_out, layer.fd_config.parallel_config.tp_group)
 
         return fused_moe_out
 

fastdeploy/rl/dynamic_weight_manager.py

@@ -220,23 +220,17 @@ class DynamicWeightManager:
         check model weights status
         """
         logger.info(f"dynamic weight manager is check model weights status! {model_weights_status.value[0]}")
-        is_stop = 0
         while model_weights_status.value[0] != ModelWeightsStatus.NORMAL:
             if model_weights_status.value[0] == ModelWeightsStatus.UPDATING:
                 logger.info("infer engine stopped! start to load new checkpoint...")
                 model_runner.update_parameters(pid)
+                while model_weights_status.value[0] != ModelWeightsStatus.NORMAL:
+                    time.sleep(0.01)
+                logger.info("finished loading new checkpoint")
             elif model_weights_status.value[0] == ModelWeightsStatus.CLEARING:
                 logger.info("infer engine stopped! start to clear checkpoint...")
                 model_runner.clear_parameters(pid)
-            while True:
+                while model_weights_status.value[0] != ModelWeightsStatus.CLEARED:
-                if model_weights_status.value[0] == ModelWeightsStatus.NORMAL:
+                    time.sleep(0.01)
-                    logger.info("finished loading new checkpoint")
-                    break
-                elif is_stop == 1 or (model_weights_status.value[0] == ModelWeightsStatus.CLEARED and is_stop == 0):
-                    if is_stop == 0:
                 logger.info("finished clearing checkpoint")
-                        is_stop = 1
+            time.sleep(0.01)
-                    time.sleep(0.001)
-                    break
-                else:
-                    time.sleep(0.001)

fastdeploy/worker/worker_process.py

@@ -270,6 +270,11 @@ class PaddleDisWorkerProc:
             create=False,
         )
 
+    def _broadcast_model_weights_signal(self, src: int, group) -> int:
+        model_weights_signal_tensor = paddle.full(shape=[1], fill_value=self.model_weights_signal[0], dtype="int32")
+        paddle.distributed.broadcast(model_weights_signal_tensor, src=src, group=group)
+        return model_weights_signal_tensor.item()
+
     def event_loop_normal(self) -> None:
         """Main event loop for Paddle Distrubuted Workers.
         TODO(gongshaotian): support remote calling of functions that control worker.
@@ -279,15 +284,19 @@ class PaddleDisWorkerProc:
         req_ids = []
         num_running_requests = 0
         local_rank = self.local_rank % self.parallel_config.tensor_parallel_size
-        self.model_weights_signal = paddle.zeros([1], dtype=paddle.int32)
+        self.model_weights_signal = np.zeros([1], dtype=np.int32)
         while True:
             if self.local_rank % self.parallel_config.tensor_parallel_size == 0:
                 if self.model_weights_status.value[0] != ModelWeightsStatus.NORMAL:
                     self.model_weights_signal[0] = int(self.model_weights_status.value[0])
                 if self.fd_config.load_config.dynamic_load_weight and self.parallel_config.enable_expert_parallel:
-                    paddle.distributed.broadcast(self.model_weights_signal, src=0, group=self.parallel_config.ep_group)
+                    self.model_weights_signal[0] = self._broadcast_model_weights_signal(
-            if self.fd_config.load_config.dynamic_load_weight:
+                        src=0, group=self.parallel_config.ep_group
-                paddle.distributed.broadcast(self.model_weights_signal, src=0, group=self.parallel_config.tp_group)
+                    )
+            if self.fd_config.load_config.dynamic_load_weight and self.parallel_config.tensor_parallel_size > 1:
+                self.model_weights_signal[0] = self._broadcast_model_weights_signal(
+                    src=0, group=self.parallel_config.tp_group
+                )
 
             self.insert_step = False
             req_dicts = None
@@ -315,7 +324,9 @@ class PaddleDisWorkerProc:
                 else:
                     paddle.distributed.barrier(self.parallel_config.tp_group)
                 if self.model_weights_signal[0] != ModelWeightsStatus.NORMAL:
-                    logger.info(f"Rank: {self.local_rank} has updated parameters.")
+                    logger.info(
+                        f"Rank: {self.local_rank} to update or clear parameters, signal is {self.model_weights_signal[0]}, [-1:clear, 1:update]"
+                    )
                     from fastdeploy.rl.dynamic_weight_manager import (
                         DynamicWeightManager,
                     )
@@ -327,6 +338,7 @@ class PaddleDisWorkerProc:
                         self.parallel_config.engine_worker_queue_port,
                     )
                     self.model_weights_signal[0] = ModelWeightsStatus.NORMAL
+                    logger.info(f"Rank: {self.local_rank} has updated or cleared parameters.")
 
             if self.exist_task_signal.value[0] == ExistTaskStatus.EXIST or self.task_queue.read_finish_flag.get() == 1:
                 logger.info(f"Rank: {self.local_rank} Detected new requests.")