[Frontend] supports deepseekv32 chat template

Signed-off-by: chaunceyjiang <[email protected]>

Author: chaunceyjiang

vllm/tokenizers/deepseekv32.py

@@ -63,7 +63,7 @@ def all_special_ids(self) -> list[int]:
 
     @property
     def bos_token_id(self) -> int:
-        raise NotImplementedError
+        return self.tokenizer.bos_token_id
 
     @property
     def eos_token_id(self) -> int:
@@ -135,13 +135,12 @@ def convert_tokens_to_string(self, tokens: list[str]) -> str:
         return self.tokenizer.convert_tokens_to_string(tokens)
 
     def decode(self, ids: list[int] | int, skip_special_tokens: bool = False) -> str:
+        skip_special_tokens = False
         return self.tokenizer.decode(ids, skip_special_tokens=skip_special_tokens)
 
     def convert_ids_to_tokens(
         self,
         ids: list[int],
         skip_special_tokens: bool = False,
     ) -> list[str]:
-        return self.tokenizer.convert_ids_to_tokens(
-            ids, skip_special_tokens=skip_special_tokens
-        )
+        return [self.decode(ids, skip_special_tokens=skip_special_tokens)]