docs: class migration from sphinx to mkdocs (engine)

Zerohertz · Zerohertz · commit b8edae13613d · 2025-05-25T19:59:51.000+09:00
Signed-off-by: Zerohertz &lt;ohg3417@gmail.com&gt;
diff --git a/vllm/engine/async_llm_engine.py b/vllm/engine/async_llm_engine.py
@@ -475,7 +475,8 @@ async def add_request_async(
             *,
             inputs: Optional[PromptType] = None,  # DEPRECATED
     ) -> None:
-        """Async version of {meth}`add_request`."""
+        """Async version of
+        [`add_request`][vllm.engine.llm_engine.LLMEngine.add_request]."""
         if inputs is not None:
             prompt = inputs
         assert prompt is not None and params is not None
@@ -582,20 +583,21 @@ async def build_guided_decoding_logits_processor_async(
 
 
 class AsyncLLMEngine(EngineClient):
-    """An asynchronous wrapper for {class}`LLMEngine`.
+    """An asynchronous wrapper for [`LLMEngine`][vllm.LLMEngine].
 
-    This class is used to wrap the {class}`LLMEngine` class to make it
-    asynchronous. It uses asyncio to create a background loop that keeps
-    processing incoming requests. The {class}`LLMEngine` is kicked by the
-    generate method when there are requests in the waiting queue. The generate
-    method yields the outputs from the {class}`LLMEngine` to the caller.
+    This class is used to wrap the [`LLMEngine`][vllm.LLMEngine] class to
+    make it asynchronous. It uses asyncio to create a background loop that keeps
+    processing incoming requests. The [`LLMEngine`][vllm.LLMEngine] is kicked
+    by the generate method when there are requests in the waiting queue. The
+    generate method yields the outputs from the [`LLMEngine`][vllm.LLMEngine]
+    to the caller.
 
     Args:
         log_requests: Whether to log the requests.
         start_engine_loop: If True, the background task to run the engine
             will be automatically started in the generate call.
-        *args: Arguments for {class}`LLMEngine`.
-        **kwargs: Arguments for {class}`LLMEngine`.
+        *args: Arguments for [`LLMEngine`][vllm.LLMEngine].
+        **kwargs: Arguments for [`LLMEngine`][vllm.LLMEngine].
     """
 
     _engine_class: Type[_AsyncLLMEngine] = _AsyncLLMEngine
@@ -985,8 +987,9 @@ async def generate(
         from the LLMEngine to the caller.
 
         Args:
-            prompt: The prompt to the LLM. See {class}`~vllm.inputs.PromptType`
-                for more details about the format of each input.
+            prompt: The prompt to the LLM. See
+                [`PromptType`][vllm.inputs.PromptType] for more details about
+                the format of each input.
             sampling_params: The sampling parameters of the request.
             request_id: The unique id of the request.
             lora_request: LoRA request to use for generation, if any.
@@ -1003,7 +1006,7 @@ async def generate(
         Details:
             - If the engine is not running, start the background loop,
               which iteratively invokes
-              {meth}`~vllm.engine.async_llm_engine.AsyncLLMEngine.engine_step`
+              [`engine_step`][vllm.engine.async_llm_engine.AsyncLLMEngine.engine_step]
               to process the waiting requests.
             - Add the request to the engine's `RequestTracker`.
               On the next background loop, this request will be sent to
@@ -1075,8 +1078,9 @@ async def encode(
         from the LLMEngine to the caller.
 
         Args:
-            prompt: The prompt to the LLM. See {class}`~vllm.inputs.PromptType`
-                for more details about the format of each input.
+            prompt: The prompt to the LLM. See
+                [`PromptType`][vllm.inputs.PromptType] for more details about
+                the format of each input.
             pooling_params: The pooling parameters of the request.
             request_id: The unique id of the request.
             lora_request: LoRA request to use for generation, if any.
diff --git a/vllm/engine/llm_engine.py b/vllm/engine/llm_engine.py
@@ -130,11 +130,11 @@ class LLMEngine:
     iteration-level scheduling and efficient memory management to maximize the
     serving throughput.
 
-    The [LLM][vllm.LLM] class wraps this class for offline batched inference
-    and the [AsyncLLMEngine][] class wraps this class for online serving.
+    The [`LLM`][vllm.LLM] class wraps this class for offline batched inference
+    and the [`AsyncLLMEngine`][vllm.engine.async_llm_engine.AsyncLLMEngine]
+    class wraps this class for online serving.
 
-    The config arguments are derived from [EngineArgs][vllm.EngineArgs]. (See
-    [engine-args][])
+    The config arguments are derived from [`EngineArgs`][vllm.EngineArgs].
 
     Args:
         vllm_config: The configuration for initializing and running vLLM.
diff --git a/vllm/engine/multiprocessing/client.py b/vllm/engine/multiprocessing/client.py
@@ -492,8 +492,9 @@ def generate(
         from the LLMEngine to the caller.
 
         Args:
-            prompt: The prompt to the LLM. See {class}`~vllm.inputs.PromptType`
-                for more details about the format of each input.
+            prompt: The prompt to the LLM. See
+                [`PromptType`][vllm.inputs.PromptType] for more details about
+                the format of each input.
             sampling_params: The sampling parameters of the request.
             request_id: The unique id of the request.
             lora_request: LoRA request to use for generation, if any.
@@ -561,8 +562,9 @@ def encode(
         from the LLMEngine to the caller.
 
         Args:
-            prompt: The prompt to the LLM. See {class}`~vllm.inputs.PromptType`
-                for more details about the format of each input.
+            prompt: The prompt to the LLM. See
+                [`PromptType`][vllm.inputs.PromptType] for more details about
+                the format of each input.
             pooling_params: The pooling parameters of the request.
             request_id: The unique id of the request.
             lora_request: LoRA request to use for generation, if any.
diff --git a/vllm/engine/multiprocessing/engine.py b/vllm/engine/multiprocessing/engine.py
@@ -42,19 +42,22 @@
 
 
 class MQLLMEngine:
-    """A multiprocessing wrapper for {class}`LLMEngine`.
+    """A multiprocessing wrapper for
+    [`LLMEngine`][vllm.engine.llm_engine.LLMEngine].
 
-    This class is used to wrap the {class}`LLMEngine` class to enable use
+    This class is used to wrap the
+    [`LLMEngine`][vllm.engine.llm_engine.LLMEngine] class to enable use
     in concurrnet manner. It runs a background loop and uses zeromq to
     receive new requests and stream outputs incrementally via ipc.
 
-    The {class}`LLMEngine` generate or encode process is kicked off when a new
-    RPCProcessRequest is received by the input_socket.
+    The [`LLMEngine`][vllm.engine.llm_engine.LLMEngine] generate or encode
+    process is kicked off when a new RPCProcessRequest is received by the
+    input_socket.
 
     The self.engine_loop checks the input_socket for new requests,
     adds them to the LLMEngine if there are any, calls the internal
-    {class}`LLMEngine.step()`, and sends the RequestOutputs back over
-    the output_socket.
+    [`LLMEngine.step()`][vllm.engine.llm_engine.LLMEngine.step], and sends
+    the RequestOutputs back over the output_socket.
 
     If use_async_sockets is set, the logic associated with reading new
     requests from the socket and sending data to the socket is passed
@@ -65,8 +68,8 @@ class MQLLMEngine:
         ipc_path: Base path for zeromq interprocess messaging
         use_async_sockets: Whether to make send/recv async with GPU
         log_requests: Whether to log the requests.
-        *args: Arguments for {class}`LLMEngine`.
-        **kwargs: Arguments for {class}`LLMEngine`.
+        *args: Arguments for [`LLMEngine`][vllm.engine.llm_engine.LLMEngine].
+        **kwargs: Arguments for [`LLMEngine`][vllm.engine.llm_engine.LLMEngine].
     """
 
     def __init__(self,
diff --git a/vllm/engine/output_processor/multi_step.py b/vllm/engine/output_processor/multi_step.py
@@ -56,8 +56,11 @@ def process_prompt_logprob(self, seq_group: SequenceGroup,
         scheduled computation.
 
         Args:
-          seq_group: the outputs are associated with this {class}`SequenceGroup`
-          outputs: the {class}`SequenceGroupOutput`s for all scheduler steps
+          seq_group: the outputs are associated with this
+              [`SequenceGroup`][vllm.sequence.SequenceGroup]
+          outputs: the
+              [`SequenceGroupOutput`][vllm.sequence.SequenceGroupOutput]s
+              for all scheduler steps
         """
         for output in outputs:
             # Concatenate single-step prompt logprob processing results.
diff --git a/vllm/engine/output_processor/single_step.py b/vllm/engine/output_processor/single_step.py
@@ -19,17 +19,21 @@
 def single_step_process_prompt_logprob(
         sg_output_proc: SequenceGroupOutputProcessor, seq_group: SequenceGroup,
         output: CompletionSequenceGroupOutput) -> None:
-    """Process prompt logprobs associated with the {class}`SequenceGroupOutput`
-    for a given step.
+    """Process prompt logprobs associated with the
+    [`SequenceGroupOutput`][vllm.sequence.SequenceGroupOutput] for a given step.
 
     Do nothing if the output has no prompt logprobs.
 
     Account for the fact that transformers do not compute first-token logprobs.
     
     Args:
-      sg_output_proc: {class}`SequenceGroupOutputProcessor` instance
-      seq_group: the output is associated with this {class}`SequenceGroup`
-      output: the {class}`SequenceGroupOutput` for a single scheduler step
+      sg_output_proc:
+          [`SequenceGroupOutputProcessor`][vllm.engine.output_processor.interfaces.SequenceGroupOutputProcessor]
+          instance
+      seq_group: the output is associated with this
+          [`SequenceGroup`][vllm.sequence.SequenceGroup]
+      output: the [`SequenceGroupOutput`][vllm.sequence.SequenceGroupOutput]
+          for a single scheduler step
     """
     prompt_logprobs = output.prompt_logprobs
 
@@ -103,8 +107,11 @@ def process_prompt_logprob(self, seq_group: SequenceGroup,
         scheduled computation.
         
         Args:
-          seq_group: the output is associated with this {class}`SequenceGroup`
-          outputs: the {class}`SequenceGroupOutput` for a single scheduler step
+          seq_group: the output is associated with this
+              [`SequenceGroup`][vllm.sequence.SequenceGroup]
+          outputs: the
+              [`SequenceGroupOutput`][vllm.sequence.SequenceGroupOutput]
+              for a single scheduler step
         """
         assert len(outputs) == 1, "Single step should only have 1 output."
         output = outputs[0]
