update docstring

lmdeploy/serve/proxy/proxy.py

@@ -422,28 +422,50 @@ async def chat_completions_v1(request: ChatCompletionRequest,
 
     The request should be a JSON object with the following fields:
     - model: model name. Available from /v1/models.
-    - messages: string prompt or chat history in OpenAI format. A example
-        for chat history is `[{"role": "user", "content":"knock knock"}]`.
+    - messages: string prompt or chat history in OpenAI format. Chat history
+        example: `[{"role": "user", "content": "hi"}]`.
     - temperature (float): to modulate the next token probability
     - top_p (float): If set to float < 1, only the smallest set of most
         probable tokens with probabilities that add up to top_p or higher
         are kept for generation.
     - n (int): How many chat completion choices to generate for each input
-        message. Only support one here.
+        message. **Only support one here**.
     - stream: whether to stream the results or not. Default to false.
-    - max_tokens (int): output token nums
+    - max_tokens (int | None): output token nums. Default to None.
     - repetition_penalty (float): The parameter for repetition penalty.
         1.0 means no penalty
     - stop (str | List[str] | None): To stop generating further
         tokens. Only accept stop words that's encoded to one token idex.
+    - response_format (Dict | None): Only pytorch backend support formatting
+        response. Examples: `{"type": "json_schema", "json_schema": {"name":
+        "test","schema": {"properties": {"name": {"type": "string"}},
+        "required": ["name"], "type": "object"}}}`
+        or `{"type": "regex_schema", "regex_schema": "call me [A-Za-z]{1,10}"}`
+    - logit_bias (Dict): Bias to logits. Only supported in pytorch engine.
+    - tools (List): A list of tools the model may call. Currently, only
+        internlm2 functions are supported as a tool. Use this to specify a
+        list of functions for which the model can generate JSON inputs.
+    - tool_choice (str | object): Controls which (if any) tool is called by
+        the model. `none` means the model will not call any tool and instead
+        generates a message. Specifying a particular tool via {"type":
+        "function", "function": {"name": "my_function"}} forces the model to
+        call that tool. `auto` or `required` will put all the tools information
+        to the model.
 
     Additional arguments supported by LMDeploy:
+    - top_k (int): The number of the highest probability vocabulary
+        tokens to keep for top-k-filtering
     - ignore_eos (bool): indicator for ignoring eos
-    - session_id (int): if not specified, will set random value
+    - skip_special_tokens (bool): Whether or not to remove special tokens
+        in the decoding. Default to be True.
+    - min_new_tokens (int): To generate at least numbers of tokens.
+    - min_p (float): Minimum token probability, which will be scaled by the
+        probability of the most likely token. It must be a value between
+        0 and 1. Typical values are in the 0.01-0.2 range, comparably
+        selective as setting `top_p` in the 0.99-0.8 range (use the
+        opposite of normal `top_p` values)
 
     Currently we do not support the following features:
-    - function_call (Users should implement this by themselves)
-    - logit_bias (not supported yet)
     - presence_penalty (replaced with repetition_penalty)
     - frequency_penalty (replaced with repetition_penalty)
     """
@@ -481,13 +503,13 @@ async def completions_v1(request: CompletionRequest,
     - model (str): model name. Available from /v1/models.
     - prompt (str): the input prompt.
     - suffix (str): The suffix that comes after a completion of inserted text.
-    - max_tokens (int): output token nums
+    - max_tokens (int): output token nums. Default to 16.
     - temperature (float): to modulate the next token probability
     - top_p (float): If set to float < 1, only the smallest set of most
         probable tokens with probabilities that add up to top_p or higher
         are kept for generation.
     - n (int): How many chat completion choices to generate for each input
-        message. Only support one here.
+        message. **Only support one here**.
     - stream: whether to stream the results or not. Default to false.
     - repetition_penalty (float): The parameter for repetition penalty.
         1.0 means no penalty
@@ -497,7 +519,8 @@ async def completions_v1(request: CompletionRequest,
 
     Additional arguments supported by LMDeploy:
     - ignore_eos (bool): indicator for ignoring eos
-    - session_id (int): if not specified, will set random value
+    - skip_special_tokens (bool): Whether or not to remove special tokens
+        in the decoding. Default to be True.
     - top_k (int): The number of the highest probability vocabulary
         tokens to keep for top-k-filtering