rename to chat_template_args

Signed-off-by: Chi McIsaac <[email protected]>

Author: qimcis

examples/multimodal/utils/chat_processor.py

@@ -160,7 +160,7 @@ async def preprocess(self, raw_request: ChatCompletionRequest) -> PreprocessResu
             continue_final_message=request.continue_final_message,
             tool_dicts=None,
             documents=request.documents,
-            chat_template_kwargs=request.chat_template_kwargs,
+            chat_template_args=request.chat_template_args,
             tool_parser=self.openai_serving.tool_parser,
             truncate_prompt_tokens=request.truncate_prompt_tokens,
             add_special_tokens=request.add_special_tokens,

lib/llm/src/entrypoint/input/batch.rs

@@ -228,7 +228,7 @@ async fn evaluate(
         inner,
         common: Default::default(),
         nvext: None,
-        chat_template_kwargs: None,
+        chat_template_args: None,
     };
     let mut stream = engine.generate(Context::new(req)).await?;
     let mut output = String::new();

lib/llm/src/entrypoint/input/text.rs

@@ -122,7 +122,7 @@ async fn main_loop(
             inner,
             common: Default::default(),
             nvext: Some(nvext),
-            chat_template_kwargs: None,
+            chat_template_args: None,
         };
 
         // Call the model

lib/llm/src/http/service/openai.rs

@@ -1324,7 +1324,7 @@ mod tests {
             },
             common: Default::default(),
             nvext: None,
-            chat_template_kwargs: None,
+            chat_template_args: None,
         };
         let result = validate_chat_completion_required_fields(&request);
         assert!(result.is_err());
@@ -1352,7 +1352,7 @@ mod tests {
             },
             common: Default::default(),
             nvext: None,
-            chat_template_kwargs: None,
+            chat_template_args: None,
         };
         let result = validate_chat_completion_required_fields(&request);
         assert!(result.is_ok());

lib/llm/src/preprocessor/prompt.rs

@@ -70,8 +70,8 @@ pub trait OAIChatLikeRequest {
 
     fn should_add_generation_prompt(&self) -> bool;
 
-    /// Optional additional kwargs to merge into the chat template context
-    fn chat_template_kwargs(&self) -> Option<HashMap<String, serde_json::Value>> {
+    /// Optional additional args to merge into the chat template context
+    fn chat_template_args(&self) -> Option<&HashMap<String, serde_json::Value>> {
         None
     }
 

lib/llm/src/preprocessor/prompt/template/oai.rs

@@ -127,8 +127,8 @@ impl OAIChatLikeRequest for NvCreateChatCompletionRequest {
         Some(TextInput::Single(String::new()))
     }
 
-    fn chat_template_kwargs(&self) -> Option<std::collections::HashMap<String, serde_json::Value>> {
-        self.chat_template_kwargs.clone()
+    fn chat_template_args(&self) -> Option<&std::collections::HashMap<String, serde_json::Value>> {
+        self.chat_template_args.as_ref()
     }
 }
 
@@ -223,9 +223,9 @@ impl OAIPromptFormatter for HfTokenizerConfigJsonFormatter {
             ..mixins
         };
 
-        // Merge any additional kwargs into the context last so they take precedence
-        let ctx = if let Some(kwargs) = req.chat_template_kwargs() {
-            let extra = Value::from_serialize(kwargs);
+        // Merge any additional args into the context last so they take precedence
+        let ctx = if let Some(args) = req.chat_template_args() {
+            let extra = Value::from_serialize(args);
             context! { ..ctx, ..extra }
         } else {
             ctx

lib/llm/src/protocols/openai/chat_completions.rs

@@ -54,9 +54,9 @@ pub struct NvCreateChatCompletionRequest {
     #[serde(skip_serializing_if = "Option::is_none")]
     pub nvext: Option<NvExt>,
 
-    /// Extra kwargs to pass to the chat template rendering context
+    /// Extra args to pass to the chat template rendering context
     #[serde(default, skip_serializing_if = "Option::is_none")]
-    pub chat_template_kwargs: Option<std::collections::HashMap<String, serde_json::Value>>,
+    pub chat_template_args: Option<std::collections::HashMap<String, serde_json::Value>>,
 }
 
 /// A response structure for unary chat completion responses, embedding OpenAI's

lib/llm/src/protocols/openai/responses.rs

@@ -187,7 +187,7 @@ impl TryFrom<NvCreateResponse> for NvCreateChatCompletionRequest {
             },
             common: Default::default(),
             nvext: resp.nvext,
-            chat_template_kwargs: None,
+            chat_template_args: None,
         })
     }
 }

lib/llm/tests/http-service.rs

@@ -768,7 +768,7 @@ async fn test_nv_custom_client() {
         inner: inner_request,
         common: Default::default(),
         nvext: None,
-        chat_template_kwargs: None,
+        chat_template_args: None,
     };
 
     let result = nv_custom_client.chat_stream(request).await;
@@ -808,7 +808,7 @@ async fn test_nv_custom_client() {
         inner: inner_request,
         common: Default::default(),
         nvext: None,
-        chat_template_kwargs: None,
+        chat_template_args: None,
     };
 
     let result = nv_custom_client.chat_stream(request).await;
@@ -849,7 +849,7 @@ async fn test_nv_custom_client() {
         inner: inner_request,
         common: Default::default(),
         nvext: None,
-        chat_template_kwargs: None,
+        chat_template_args: None,
     };
 
     let result = nv_custom_client

lib/llm/tests/preprocessor.rs

@@ -270,7 +270,7 @@ impl Request {
             inner,
             common: Default::default(),
             nvext: None,
-            chat_template_kwargs: None,
+            chat_template_args: None,
         }
     }
 }