[0.9.1] fix error info

vllm_ascend/platform.py

@@ -27,6 +27,7 @@
 from vllm.logger import logger
 from vllm.platforms import Platform, PlatformEnum
 
+import vllm_ascend.envs as envs_ascend
 from vllm_ascend.ascend_config import check_ascend_config, init_ascend_config
 from vllm_ascend.utils import (ASCEND_QUATIZATION_METHOD,
                                check_torchair_cache_exist,
@@ -169,6 +170,10 @@ def check_and_update_config(cls, vllm_config: VllmConfig) -> None:
 
         check_ascend_config(vllm_config, enforce_eager)
 
+        if vllm_config.speculative_config and envs_ascend.VLLM_ASCEND_ENABLE_DBO:
+            raise ValueError(
+                "DBO and mtp can't work at the same time. Please `export VLLM_ASCEND_ENABLE_DBO=0`"
+            )
         if enforce_eager or compilation_config.level == CompilationLevel.NO_COMPILATION:
             logger.info("Compilation disabled, using eager mode by default")
             compilation_config.level = CompilationLevel.NO_COMPILATION
@@ -194,7 +199,7 @@ def check_and_update_config(cls, vllm_config: VllmConfig) -> None:
             logger.warning(
                 "Ray distributed executor backend is not compatible with ACL Graph mode "
                 "right now. Setting level to NO_COMPILATION")
-            compilation_config.level = CompilationLevel.NO_COMPILATION
+            compilation_config.level = CompilationLevel.NfvO_COMPILATION
         else:
             logger.info(
                 "PIECEWISE compilation enabled on NPU. use_inductor not supported - "

vllm_ascend/worker/model_runner_v1.py

@@ -221,9 +221,6 @@ def __init__(self, vllm_config: VllmConfig, device: torch.device):
         self.spec_token_num = 0
         self.decode_token_per_req = 1
         if self.speculative_config:
-            if envs_ascend.VLLM_ASCEND_ENABLE_DBO:
-                raise NotImplementedError(
-                    "DBO and mtp can't work at the same currently")
             self.use_spec_decode = True
             self.spec_token_num = self.speculative_config.num_speculative_tokens
             assert self.spec_token_num > 0