fix typos

custom_ops/xpu_ops/src/plugin/src/kernel/kunlun3cpp/mtp_kernel/speculate_update_repeat_times.xpu

@@ -190,7 +190,7 @@ __device__ void speculate_update_repeat_times_optimized(
           buffer_ptr_pre_ids.toggle();
         }
       }
-      // each core loads all the needed pre_ids into lm without mfence inbetween
+      // each core loads all the needed pre_ids into lm without mfence in between
       // according to the index recorded by previous iteration
       else {
         int cnt = -1;

fastdeploy/engine/common_engine.py

@@ -508,7 +508,7 @@ def _insert_task_to_worker(self):
                 main_process_metrics.num_requests_waiting.dec(len(tasks))
                 main_process_metrics.num_requests_running.inc(len(tasks))
             except Exception as e:
-                err_msg = f"Error happend while insert task to engine: {e}, {traceback.format_exc()!s}."
+                err_msg = f"Error happened while insert task to engine: {e}, {traceback.format_exc()!s}."
                 llm_logger.error(err_msg)
 
     def _scheduler_task_to_worker_v1(self):
@@ -560,7 +560,7 @@ def _fetch_request():
                     time.sleep(0.005)
 
             except Exception as e:
-                err_msg = "Error happend while insert task to engine: {}, {}.".format(e, str(traceback.format_exc()))
+                err_msg = "Error happened while insert task to engine: {}, {}.".format(e, str(traceback.format_exc()))
                 llm_logger.error(err_msg)
 
     def start_zmq_service(self, api_server_pid=None):
@@ -642,7 +642,7 @@ def _insert_zmq_task_to_scheduler(self):
                     self.zmq_server.send_multipart(request_id, [error_result])
             except Exception as e:
                 llm_logger.error(
-                    f"Error happend while receiving new request from zmq, details={e}, "
+                    f"Error happened while receiving new request from zmq, details={e}, "
                     f"traceback={traceback.format_exc()}"
                 )
 
@@ -660,7 +660,7 @@ def _zmq_send_generated_tokens(self):
                     self.zmq_server.send_multipart(request_id, contents)
 
             except Exception as e:
-                llm_logger.error(f"Unexcepted error happend: {e}, {traceback.format_exc()!s}")
+                llm_logger.error(f"Unexcepted error happened: {e}, {traceback.format_exc()!s}")
 
     def split_mode_get_tasks(self):
         """

fastdeploy/engine/engine.py

@@ -563,7 +563,7 @@ def generate(self, prompts, stream):
         try:
             req_id = self._format_and_add_data(prompts)
         except Exception as e:
-            llm_logger.error(f"Error happend while adding request, details={e}, {str(traceback.format_exc())}")
+            llm_logger.error(f"Error happened while adding request, details={e}, {str(traceback.format_exc())}")
             raise EngineError(str(e), error_code=400)
 
         # Get the result of the current request

fastdeploy/entrypoints/engine_client.py

@@ -204,8 +204,8 @@ async def add_requests(self, task):
             f"preprocess time cost {preprocess_cost_time}"
         )
 
-        self.vaild_parameters(task)
-        api_server_logger.debug(f"Recieve task: {task}")
+        self.valid_parameters(task)
+        api_server_logger.debug(f"Receive task: {task}")
         try:
             if not self.enable_mm:
                 self.zmq_client.send_json(task)
@@ -215,7 +215,7 @@ async def add_requests(self, task):
             api_server_logger.error(f"zmq_client send task error: {e}, {str(traceback.format_exc())}")
             raise EngineError(str(e), error_code=400)
 
-    def vaild_parameters(self, data):
+    def valid_parameters(self, data):
         """
         Validate stream options
         """

fastdeploy/entrypoints/llm.py

@@ -125,7 +125,7 @@ def _receive_output(self):
                                 continue
                             self.req_output[request_id].add(result)
             except Exception as e:
-                llm_logger.error(f"Unexcepted error happend: {e}, {traceback.format_exc()!s}")
+                llm_logger.error(f"Unexcepted error happened: {e}, {traceback.format_exc()!s}")
 
     def generate(
         self,

fastdeploy/model_executor/layers/attention/moba_attention_backend.py

@@ -124,7 +124,7 @@ def get_kv_cache_shape(
         kv_cache_quant_type: str = None,
     ):
         """
-        Caculate kv cache shape
+        Calculate kv cache shape
         """
         if kv_cache_quant_type is not None and kv_cache_quant_type == "int4_zp":
             return (

fastdeploy/model_executor/layers/lm_head.py

@@ -56,7 +56,7 @@ def __init__(
             embedding_dim (int): size of hidden state.
             prefix (str): The name of current layer. Defaults to "".
             with_bias (bool): whether to have bias. Default: False.
-            dtype (str): The dtype of weight. Defalut: None.
+            dtype (str): The dtype of weight. Default: None.
         """
         super(ParallelLMHead, self).__init__()
         self.weight_key: str = prefix + ".weight"

fastdeploy/model_executor/layers/sample/sampler.py

@@ -364,7 +364,7 @@ def forward_cuda(
         )
         if sampling_metadata.enable_early_stop:
             # will set the stop batch in stop_flags
-            assert sampling_metadata.stop_flags is not None, "need stop_flags for eary stop"
+            assert sampling_metadata.stop_flags is not None, "need stop_flags for early stop"
             self.early_stopper.process(probs, next_tokens, sampling_metadata.stop_flags)
 
         sampler_output = SamplerOutput(

fastdeploy/model_executor/ops/triton_ops/triton_utils.py

@@ -683,7 +683,7 @@ def decorator(*args, **kwargs):
             op_dict = {"op_name": op_name, "reset_zero_when_tune": ""}
             op_dict["triton_kernel_args"] = ",".join(modified_arg_exclude_constexpr)
             op_dict["key"] = ",".join(self.key_args)
-            # when tunning, we need to reset the out to zero.
+            # when tuning, we need to reset the out to zero.
             if "reset_zero_when_tune" in other_config.keys():
                 op_dict["reset_zero_when_tune"] = other_config["reset_zero_when_tune"]
 

fastdeploy/output/token_processor.py

@@ -178,7 +178,7 @@ def process_sampling_results(self):
                                 )
 
                     except Exception as e:
-                        print(f"Recieve message error: {e}")
+                        print(f"Receive message error: {e}")
                         continue
                 else:
                     is_blocking = True

fastdeploy/rl/dynamic_weight_manager.py

@@ -105,7 +105,7 @@ def _update_ipc(self):
 
     def clear_parameters(self, pid: int = 0) -> None:
         """Clear all model parameters and free memory."""
-        logger.info("start clear paramaters")
+        logger.info("start clear parameters")
         paddle.device.cuda.empty_cache()
         for param in self.model.state_dict().values():
             param._clear_data()

fastdeploy/rl/rollout_model.py

@@ -146,7 +146,7 @@ def name(self) -> str:
         return "Ernie4_5_MoeForCausalLMRL"
 
     def get_name_mappings_to_training(self, trainer_degree=None) -> Dict[str, str]:
-        """Generate mapping between inference and training parameter for RL(donot delete!)."""
+        """Generate mapping between inference and training parameter for RL(do not delete!)."""
         if self._mappings_built:
             return self.infer_to_train_mapping
 
@@ -225,7 +225,7 @@ def name(self) -> str:
         return "Ernie4_5_VLMoeForConditionalGenerationRL"
 
     def get_name_mappings_to_training(self, trainer_degree=None) -> Dict[str, str]:
-        """Generate mapping between inference and training parameter for RL(donot delete!)."""
+        """Generate mapping between inference and training parameter for RL(do not delete!)."""
         if self._mappings_built:
             return self.infer_to_train_mapping
 
@@ -331,7 +331,7 @@ def name(self) -> str:
         return "Qwen2ForCausalLMRL"
 
     def get_name_mappings_to_training(self, trainer_degree=None) -> Dict[str, str]:
-        """Generate mapping between inference and training parameter for RL(donot delete!)."""
+        """Generate mapping between inference and training parameter for RL(do not delete!)."""
         if self._mappings_built:
             return self.infer_to_train_mapping
 
@@ -380,7 +380,7 @@ def name(self) -> str:
         return "Qwen3MoeForCausalLMRL"
 
     def get_name_mappings_to_training(self, trainer_degree=None) -> Dict[str, str]:
-        """Generate mapping between inference and training parameter for RL(donot delete!)."""
+        """Generate mapping between inference and training parameter for RL(do not delete!)."""
         if self._mappings_built:
             return self.infer_to_train_mapping
 

fastdeploy/scheduler/global_scheduler.py

@@ -648,7 +648,7 @@ def _put_results_worker(self, tasks: List[Task]):
                 stolen_responses[response_queue_name].append(response.serialize())
                 continue
 
-            scheduler_logger.error(f"Scheduler has recieved a non-existent response from engine: {[response]}")
+            scheduler_logger.error(f"Scheduler has received a non-existent response from engine: {[response]}")
 
         with self.mutex:
             for request_id, responses in local_responses.items():

fastdeploy/worker/dcu_worker.py

@@ -49,7 +49,7 @@ def init_device(self):
         """
         self.max_chips_per_node = 8
         if self.device_config.device_type == "cuda" and paddle.device.is_compiled_with_cuda():
-            # Set evironment variable
+            # Set environment variable
             self.device_ids = self.parallel_config.device_ids.split(",")
             self.device = f"gpu:{self.local_rank % self.max_chips_per_node}"
             paddle.device.set_device(self.device)