Support limit thinking lengths (#4070)

Co-authored-by: K11OntheBoat <“ruianmaidanglao@163.com”>

fastdeploy/config.py

@@ -130,6 +130,7 @@ class ModelConfig:
         self.quantization = None
         self.pad_token_id: int = -1
         self.eos_tokens_lens: int = 2
+        self.think_end_id = None
         self.lm_head_fp32: bool = False
         self.model_format = "auto"
         self.partial_rotary_factor: float = 1.0

fastdeploy/entrypoints/engine_client.py

@@ -177,8 +177,6 @@ class EngineClient:
             task["prompt_token_ids_len"] = len(task["prompt_token_ids"])
             input_ids_len = task["prompt_token_ids_len"]
             task["max_tokens"] = min(self.max_model_len - input_ids_len, task.get("max_tokens"))
-            if task.get("reasoning_max_tokens", None) is None:
-                task["reasoning_max_tokens"] = max(int(task["max_tokens"] * 0.8), 1)
             min_tokens = task.get("min_tokens", 1)
             if "messages" in task:
                 del task["messages"]

fastdeploy/input/ernie4_5_vl_processor/ernie4_5_vl_processor.py

@@ -255,6 +255,10 @@ class Ernie4_5_VLProcessor(Ernie4_5Processor):
             request["prompt_token_ids"] = request["prompt_token_ids"][: max_model_len - 1]
         if request.get("max_tokens") is None:
             request["max_tokens"] = max(1, max_model_len - len(request["prompt_token_ids"]))
+        else:
+            request["max_tokens"] = min(max_model_len - len(request["prompt_token_ids"]), request["max_tokens"])
+        if request.get("reasoning_max_tokens") is None:
+            request["reasoning_max_tokens"] = max(int(request["max_tokens"] * 0.8), 1)
         data_processor_logger.info(f"Processed request {request}")
 
         return request

fastdeploy/model_executor/pre_and_post_process.py

@@ -166,7 +166,7 @@ def post_process_normal(
 ) -> ModelRunnerOutput:
     """Post-processing steps after completing a single token generation."""
     # handle vl:
-    if model_output.enable_thinking:
+    if model_output.enable_thinking and model_output.think_end_id is not None:
         exists_think_end = sampler_output.sampled_token_ids == model_output.think_end_id
         paddle.assign(
             paddle.where(

fastdeploy/worker/gpu_model_runner.py

@@ -265,15 +265,21 @@ class GPUModelRunner(ModelRunnerBase):
                     else:
                         position_ids = None
 
-                    enable_thinking = request.get("enable_thinking", True)
-                    enable_thinking = enable_thinking if enable_thinking is not None else True
-                    self.share_inputs["enable_thinking"][:] = enable_thinking
-                    self.share_inputs["need_think_end"][idx : idx + 1, :] = 1 if enable_thinking else 0
-                    self.share_inputs["reasoning_index"][idx : idx + 1, :] = request.get("reasoning_max_tokens", 2048)
                     self.share_inputs["rope_emb"][idx : idx + 1, :] = self.prepare_rope3d(
                         position_ids, request.get("max_tokens", 2048)
                     )
 
+                if request.get("enable_thinking", False) and request.get("reasoning_max_tokens") is not None:
+                    # Enable thinking
+                    self.share_inputs["enable_thinking"][:] = True
+                    self.share_inputs["need_think_end"][idx : idx + 1, :] = 1
+                    self.share_inputs["reasoning_index"][idx : idx + 1, :] = request.get("reasoning_max_tokens")
+                else:
+                    # Disable thinking
+                    self.share_inputs["enable_thinking"][:] = False
+                    self.share_inputs["need_think_end"][idx : idx + 1, :] = 0
+                    self.share_inputs["reasoning_index"][idx : idx + 1, :] = 0
+
                 if isinstance(request.prompt_token_ids, np.ndarray):
                     prompt_token_ids = request.prompt_token_ids.tolist()
                 else:
@@ -495,16 +501,22 @@ class GPUModelRunner(ModelRunnerBase):
                     self.share_inputs["prompt_lens"][idx : idx + 1] = length
 
                 if self.enable_mm:
-                    enable_thinking = request.get("enable_thinking", True)
-                    enable_thinking = enable_thinking if enable_thinking is not None else True
-                    self.share_inputs["enable_thinking"][:] = enable_thinking
-                    self.share_inputs["need_think_end"][idx : idx + 1, :] = 1 if enable_thinking else 0
-                    self.share_inputs["reasoning_index"][idx : idx + 1, :] = request.get("reasoning_max_tokens", 2048)
                     self.share_inputs["rope_emb"][idx : idx + 1, :] = self.prepare_rope3d(
                         position_ids, request.get("max_tokens", 2048)
                     )
                     self.share_inputs["seq_lens_decoder"][idx : idx + 1] = 0
 
+                if request.get("enable_thinking", False) and request.get("reasoning_max_tokens") is not None:
+                    # Enable thinking
+                    self.share_inputs["enable_thinking"][:] = True
+                    self.share_inputs["need_think_end"][idx : idx + 1, :] = 1
+                    self.share_inputs["reasoning_index"][idx : idx + 1, :] = request.get("reasoning_max_tokens")
+                else:
+                    # Disable thinking
+                    self.share_inputs["enable_thinking"][:] = False
+                    self.share_inputs["need_think_end"][idx : idx + 1, :] = 0
+                    self.share_inputs["reasoning_index"][idx : idx + 1, :] = 0
+
             def get_attr_from_request(request, attr, default_value=None):
                 res = request.get(attr, default_value)
                 if res is not None:
@@ -735,6 +747,11 @@ class GPUModelRunner(ModelRunnerBase):
         # Initialize rotary position embedding
         tmp_position_ids = paddle.arange(self.parallel_config.max_model_len).reshape((1, -1))
 
+        # Initialize thinking related buffers
+        self.share_inputs["need_think_end"] = paddle.full(shape=[max_num_seqs, 1], fill_value=0, dtype="int32")
+        self.share_inputs["enable_thinking"] = paddle.full(shape=[1], fill_value=False, dtype="bool")
+        self.share_inputs["reasoning_index"] = paddle.full(shape=[max_num_seqs, 1], fill_value=0, dtype="int32")
+
         # TODO(gongshaotian): move to models
         if not self.enable_mm:
             self.share_inputs["rope_emb"] = get_rope(
@@ -827,11 +844,6 @@ class GPUModelRunner(ModelRunnerBase):
                 dtype="float32",
             )
             self.share_inputs["image_features"] = None
-            self.share_inputs["need_think_end"] = paddle.full(shape=[max_num_seqs, 1], fill_value=0, dtype="int32")
-            self.share_inputs["enable_thinking"] = paddle.full(
-                shape=[1], fill_value=("ernie" in self.model_config.model_type), dtype="bool"
-            )
-            self.share_inputs["reasoning_index"] = paddle.full(shape=[max_num_seqs, 1], fill_value=0, dtype="int32")
 
     def _prepare_inputs(self) -> None:
         """Prepare the model inputs"""
@@ -1220,10 +1232,10 @@ class GPUModelRunner(ModelRunnerBase):
                 ),
                 accept_tokens=(self.share_inputs["accept_tokens"] if self.speculative_decoding else None),
                 accept_num=(self.share_inputs["accept_num"] if self.speculative_decoding else None),
-                enable_thinking=(self.share_inputs["enable_thinking"] if self.enable_mm else None),
+                enable_thinking=self.share_inputs["enable_thinking"],
-                think_end_id=(getattr(self.model_config, "think_end_id", -1) if self.enable_mm else -1),
+                think_end_id=self.model_config.think_end_id,
-                need_think_end=(self.share_inputs["need_think_end"] if self.enable_mm else None),
+                need_think_end=self.share_inputs["need_think_end"],
-                reasoning_index=(self.share_inputs["reasoning_index"] if self.enable_mm else None),
+                reasoning_index=self.share_inputs["reasoning_index"],
                 stop_token_ids=self.share_inputs["stop_seqs"],
                 stop_seqs_len=self.share_inputs["stop_seqs_len"],
             )
@@ -1515,10 +1527,10 @@ class GPUModelRunner(ModelRunnerBase):
             ),
             accept_tokens=(self.share_inputs["accept_tokens"] if self.speculative_decoding else None),
             accept_num=(self.share_inputs["accept_num"] if self.speculative_decoding else None),
-            enable_thinking=(self.share_inputs["enable_thinking"] if self.enable_mm else None),
+            enable_thinking=self.share_inputs["enable_thinking"],
-            think_end_id=(getattr(self.model_config, "think_end_id", -1) if self.enable_mm else -1),
+            think_end_id=self.model_config.think_end_id,
-            need_think_end=(self.share_inputs["need_think_end"][:num_running_requests] if self.enable_mm else None),
+            need_think_end=self.share_inputs["need_think_end"][:num_running_requests],
-            reasoning_index=(self.share_inputs["reasoning_index"][:num_running_requests] if self.enable_mm else None),
+            reasoning_index=self.share_inputs["reasoning_index"][:num_running_requests],
             stop_token_ids=self.share_inputs["stop_seqs"],
             stop_seqs_len=self.share_inputs["stop_seqs_len"],
         )

fastdeploy/worker/worker_process.py

@@ -129,6 +129,28 @@ def update_fd_config_for_mm(fd_config: FDConfig) -> None:
         fd_config.model_config.sequence_parallel = fd_config.parallel_config.sequence_parallel
 
 
+def update_think_end_id_for_ernie(fd_config: FDConfig) -> None:
+    """
+    Updates the think_end_id in the model config. Uses the ID of '</think>'
+    if it exists, otherwise defaults to None.
+    """
+    is_ernie = ErnieArchitectures.contains_ernie_arch(fd_config.model_config.architectures)
+    if current_platform.is_cuda() and is_ernie:
+        tokenizer = Ernie4_5Tokenizer.from_pretrained(
+            fd_config.model_config.model,
+            model_max_length=fd_config.parallel_config.max_model_len,
+            padding_side="right",
+            use_fast=False,
+        )
+
+        vocab = tokenizer.get_vocab()
+        fd_config.model_config.think_end_id = vocab.get("</think>", None)
+        if fd_config.model_config.think_end_id is not None:
+            logger.info(f"Get think_end_id {fd_config.model_config.think_end_id} from vocab.")
+        else:
+            logger.info("No </think> token found in vocabulary, the model can not do reasoning.")
+
+
 class PaddleDisWorkerProc:
     """
     Paddle Distributed wrapper for fastdeploy.worker.Worker,
@@ -771,6 +793,7 @@ def initialize_fd_config(args, ranks: int = 1, local_rank: int = 0) -> FDConfig:
         plas_attention_config=plas_attention_config,
     )
     update_fd_config_for_mm(fd_config)
+    update_think_end_id_for_ernie(fd_config)
 
     return fd_config
 

tests/ci_use/EB_VL_Lite/test_EB_VL_Lite_serving.py

@@ -580,3 +580,63 @@ def test_profile_reset_block_num():
         f"Reset total_block_num {actual_value} 与 baseline {baseline} diff需要在5%以内"
         f"Allowed range: [{lower_bound:.1f}, {upper_bound:.1f}]"
     )
+
+
+def test_thinking_logic_flag(openai_client, capsys):
+    """
+    Test the interaction between token calculation logic and conditional thinking.
+    This test covers:
+    1. Default max_tokens calculation when not provided.
+    2. Capping of max_tokens when it exceeds model limits.
+    3. Default reasoning_max_tokens calculation when not provided.
+    4. Activation of thinking based on the final state of reasoning_max_tokens.
+    """
+
+    response_case_1 = openai_client.chat.completions.create(
+        model="default",
+        messages=[{"role": "user", "content": "Explain gravity briefly."}],
+        temperature=1,
+        stream=False,
+        extra_body={
+            "chat_template_kwargs": {"enable_thinking": True},
+        },
+    )
+    assert response_case_1.choices[0].message.reasoning_content is not None
+
+    response_case_2 = openai_client.chat.completions.create(
+        model="default",
+        messages=[{"role": "user", "content": "Explain gravity in a way that a five-year-old child can understand."}],
+        temperature=1,
+        stream=False,
+        max_tokens=20,
+        extra_body={
+            "chat_template_kwargs": {"enable_thinking": True},
+            "reasoning_max_tokens": 5,
+        },
+    )
+    assert response_case_2.choices[0].message.reasoning_content is not None
+
+    response_case_3 = openai_client.chat.completions.create(
+        model="default",
+        messages=[{"role": "user", "content": "Explain gravity in a way that a five-year-old child can understand."}],
+        temperature=1,
+        stream=False,
+        max_tokens=20,
+        extra_body={
+            "chat_template_kwargs": {"enable_thinking": True},
+            "reasoning_max_tokens": None,
+        },
+    )
+    assert response_case_3.choices[0].message.reasoning_content is not None
+
+    response_case_4 = openai_client.chat.completions.create(
+        model="default",
+        messages=[{"role": "user", "content": "Explain gravity in a way that a five-year-old child can understand."}],
+        temperature=1,
+        stream=False,
+        max_tokens=20,
+        extra_body={
+            "chat_template_kwargs": {"enable_thinking": False},
+        },
+    )
+    assert response_case_4.choices[0].message.reasoning_content is None

tests/e2e/test_EB_VL_Lite_serving.py

@@ -592,3 +592,63 @@ def test_profile_reset_block_num():
         f"Reset total_block_num {actual_value} 与 baseline {baseline} diff需要在5%以内"
         f"Allowed range: [{lower_bound:.1f}, {upper_bound:.1f}]"
     )
+
+
+def test_thinking_logic_flag(openai_client, capsys):
+    """
+    Test the interaction between token calculation logic and conditional thinking.
+    This test covers:
+    1. Default max_tokens calculation when not provided.
+    2. Capping of max_tokens when it exceeds model limits.
+    3. Default reasoning_max_tokens calculation when not provided.
+    4. Activation of thinking based on the final state of reasoning_max_tokens.
+    """
+
+    response_case_1 = openai_client.chat.completions.create(
+        model="default",
+        messages=[{"role": "user", "content": "Explain gravity briefly."}],
+        temperature=1,
+        stream=False,
+        extra_body={
+            "chat_template_kwargs": {"enable_thinking": True},
+        },
+    )
+    assert response_case_1.choices[0].message.reasoning_content is not None
+
+    response_case_2 = openai_client.chat.completions.create(
+        model="default",
+        messages=[{"role": "user", "content": "Explain gravity in a way that a five-year-old child can understand."}],
+        temperature=1,
+        stream=False,
+        max_tokens=20,
+        extra_body={
+            "chat_template_kwargs": {"enable_thinking": True},
+            "reasoning_max_tokens": 5,
+        },
+    )
+    assert response_case_2.choices[0].message.reasoning_content is not None
+
+    response_case_3 = openai_client.chat.completions.create(
+        model="default",
+        messages=[{"role": "user", "content": "Explain gravity in a way that a five-year-old child can understand."}],
+        temperature=1,
+        stream=False,
+        max_tokens=20,
+        extra_body={
+            "chat_template_kwargs": {"enable_thinking": True},
+            "reasoning_max_tokens": None,
+        },
+    )
+    assert response_case_3.choices[0].message.reasoning_content is not None
+
+    response_case_4 = openai_client.chat.completions.create(
+        model="default",
+        messages=[{"role": "user", "content": "Explain gravity in a way that a five-year-old child can understand."}],
+        temperature=1,
+        stream=False,
+        max_tokens=20,
+        extra_body={
+            "chat_template_kwargs": {"enable_thinking": False},
+        },
+    )
+    assert response_case_4.choices[0].message.reasoning_content is None