【FIX】Change the name of sparse attn from moba to plas (#4006) (#4076)

* 【FIX】Change the name of sparse attn from moba to plas (#4006)

* 更新文档

* 【docs】 update readme (#4000)

* 更新文档

* update readme

* update docs

* 【FIX】Change the name of sparse attn from moba to plas (#3845)

* 更新文档

* 更新文档

* 更新文档

* 更新文档

* 修改moba为plas

* code style

* update ci

* code style

* update ci

* code style

---------

Co-authored-by: Jiang-Jia-Jun <163579578+Jiang-Jia-Jun@users.noreply.github.com>

* fix max_num_seqs

* fix test load attn

---------

Co-authored-by: Jiang-Jia-Jun <163579578+Jiang-Jia-Jun@users.noreply.github.com>

fastdeploy/engine/args_utils.py

@@ -30,9 +30,9 @@ from fastdeploy.config import (
     FDConfig,
     GraphOptimizationConfig,
     LoadConfig,
-    MobaAttentionConfig,
     ModelConfig,
     ParallelConfig,
+    PlasAttentionConfig,
     PoolerConfig,
     RunnerOption,
     SpeculativeConfig,
@@ -361,9 +361,9 @@ class EngineArgs:
     """
     Configuration for graph optimization backend execution.
     """
-    moba_attention_config: Optional[Dict[str, Any]] = None
+    plas_attention_config: Optional[Dict[str, Any]] = None
     """
-    Configuration for moba attention.
+    Configuration for plas attention.
     """
 
     enable_logprob: bool = False
@@ -601,9 +601,9 @@ class EngineArgs:
             help="",
         )
         model_group.add_argument(
-            "--moba-attention-config",
+            "--plas-attention-config",
             type=json.loads,
-            default=EngineArgs.moba_attention_config,
+            default=EngineArgs.plas_attention_config,
             help="",
         )
         model_group.add_argument(
@@ -993,17 +993,17 @@ class EngineArgs:
                 graph_optimization_args[k] = v
         return GraphOptimizationConfig(graph_optimization_args)
 
-    def create_moba_attention_config(self) -> MobaAttentionConfig:
+    def create_plas_attention_config(self) -> PlasAttentionConfig:
         """
-        Create and retuan a MobaAttentionConfig object based on the current settings.
+        Create and retuan a PlasAttentionConfig object based on the current settings.
         """
         attention_args = asdict(self)
-        if self.moba_attention_config is not None:
-            for k, v in self.moba_attention_config.items():
+        if self.plas_attention_config is not None:
+            for k, v in self.plas_attention_config.items():
                 attention_args[k] = v
-            return MobaAttentionConfig(attention_args)
+            return PlasAttentionConfig(attention_args)
         else:
-            return MobaAttentionConfig(None)
+            return PlasAttentionConfig(None)
 
     def create_early_stop_config(self) -> EarlyStopConfig:
         """
@@ -1064,7 +1064,7 @@ class EngineArgs:
         scheduler_cfg = self.create_scheduler_config()
         graph_opt_cfg = self.create_graph_optimization_config()
         graph_opt_cfg.update_use_cudagraph(self.use_cudagraph)
-        moba_attention_config = self.create_moba_attention_config()
+        plas_attention_config = self.create_plas_attention_config()
 
         early_stop_cfg = self.create_early_stop_config()
         early_stop_cfg.update_enable_early_stop(self.enable_early_stop)
@@ -1093,7 +1093,7 @@ class EngineArgs:
             max_long_partial_prefills=self.max_long_partial_prefills,
             long_prefill_token_threshold=self.long_prefill_token_threshold,
             graph_opt_config=graph_opt_cfg,
-            moba_attention_config=moba_attention_config,
+            plas_attention_config=plas_attention_config,
             guided_decoding_backend=self.guided_decoding_backend,
             disable_any_whitespace=self.guided_decoding_disable_any_whitespace,
             early_stop_config=early_stop_cfg,

fastdeploy/engine/engine.py

@@ -501,7 +501,7 @@ class LLMEngine:
             f" --early_stop_config '{self.cfg.early_stop_config.to_json_string()}'"
             f" --reasoning_parser {self.cfg.reasoning_parser}"
             f" --load_choices {self.cfg.load_config.load_choices}"
-            f" --moba_attention_config '{self.cfg.moba_attention_config.to_json_string()}'"
+            f" --plas_attention_config '{self.cfg.plas_attention_config.to_json_string()}'"
             f" --ips {ips}"
             f" --cache-transfer-protocol {self.cfg.cache_config.cache_transfer_protocol}"
             f" --runner {self.cfg.model_config.runner}"