[Feature] mm and thinking model support structred output (#2749)

* mm support structured output

* update code

* update code

* update format

* update code

* update code

* add enable_thinking default

* update code

* add structured_outputs test case

* add ci install xgrammar

* add ci timeout time

* update test for structured_outputs

* update code

* add error traceback info

* update error msg

* update structred output code

* update code

* update code

* update config

* update torch version

---------

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

fastdeploy/config.py

@@ -127,12 +127,13 @@ class ModelConfig:
         self.redundant_experts_num = 0
         self.seed = 0
         self.quantization = None
+        self.reasoning_parser = None
         self.pad_token_id: int = -1
         self.eos_tokens_lens: int = 2
         self.lm_head_fp32: bool = False
         self.model_format = "auto"
         for key, value in args.items():
-            if hasattr(self, key):
+            if hasattr(self, key) and value != "None":
                 setattr(self, key, value)
 
         assert self.model != ""
@@ -1249,7 +1250,8 @@ class FDConfig:
         self.cache_config.max_block_num_per_seq = int(self.max_model_len // self.cache_config.block_size)
 
         if self.guided_decoding_backend == "auto":
-            if self.model_config.enable_mm:
+            if current_platform.is_xpu() or self.speculative_config.method is not None:
+                logger.warning("Speculative Decoding and XPU currently do not support Guided decoding, set off.")
                 self.guided_decoding_backend = "off"
             else:
                 self.guided_decoding_backend = "xgrammar"
@@ -1319,12 +1321,10 @@ class FDConfig:
             ], f"Only support xgrammar、auto guided decoding backend, but got {self.guided_decoding_backend}."
 
             if self.guided_decoding_backend != "off":
-                # TODO: mm support guided_decoding
-                assert (
-                    self.model_config.enable_mm is False
-                ), "Multimodal model currently do not support guided_decoding"
-
                 # TODO: speculative decoding support guided_decoding
+                assert (
+                    self.speculative_config.method is None
+                ), "speculative decoding currently do not support guided_decoding"
 
                 # TODO: xpu support guided_decoding
                 assert not current_platform.is_xpu(), "XPU currently do not support guided_decoding"
@@ -1335,6 +1335,7 @@ class FDConfig:
                     raise Exception(
                         f"import XGrammar failed, please install XGrammar use `pip install xgrammar==0.1.19`. \n\t {e}"
                     )
+
         if self.scheduler_config is not None:
             self.scheduler_config.check()