diff --git a/rust/src/server/src/routes/tests.rs b/rust/src/server/src/routes/tests.rs index c6c05e63b1..d1e9a55d16 100644 --- a/rust/src/server/src/routes/tests.rs +++ b/rust/src/server/src/routes/tests.rs @@ -1374,6 +1374,40 @@ async fn render_chat_returns_generate_request_with_header_request_id() { assert!(json.get("mm_features").is_none()); } +#[tokio::test] +async fn render_chat_omits_unset_sampling_params() { + let mut app = test_render_app(); + let response = app + .call( + Request::builder() + .method("POST") + .uri("/v1/chat/completions/render") + .header("content-type", "application/json") + .body(Body::from( + json!({ + "model": "render-model", + "messages": [{"role": "user", "content": "hello"}], + "max_completion_tokens": 8 + }) + .to_string(), + )) + .expect("build request"), + ) + .await + .expect("call app"); + + assert_eq!(response.status(), StatusCode::OK); + let body = to_bytes(response.into_body(), usize::MAX).await.expect("read body"); + let json: serde_json::Value = serde_json::from_slice(&body).expect("decode json"); + + let sampling_params = json["sampling_params"].as_object().expect("sampling_params object"); + assert!( + sampling_params.values().all(|value| !value.is_null()), + "unexpected null in {sampling_params:?}" + ); + assert_eq!(sampling_params["max_tokens"], 8); +} + #[tokio::test] async fn render_chat_uses_processor_owned_parser_selections() { let mut app = test_render_app_with_parser_selections( diff --git a/rust/src/text/src/request.rs b/rust/src/text/src/request.rs index b00d8f3b2f..c3efcde496 100644 --- a/rust/src/text/src/request.rs +++ b/rust/src/text/src/request.rs @@ -65,6 +65,7 @@ impl Default for Prompt { /// /// Original Python definition: /// +#[serde_with::skip_serializing_none] #[derive(Debug, Clone, PartialEq, Serialize, Deserialize)] #[serde(default)] pub struct SamplingParams { @@ -320,4 +321,23 @@ mod tests { assert!(err.to_string().contains("stop strings cannot be empty")); assert!(err.to_string().contains("req-empty-stop")); } + + #[test] + fn sampling_params_serialization_omits_unset_fields() { + let params = SamplingParams { + max_tokens: Some(64), + ..Default::default() + }; + + let value = serde_json::to_value(¶ms).unwrap(); + let obj = value.as_object().unwrap(); + assert!( + obj.values().all(|v| !v.is_null()), + "unexpected null in {value}" + ); + assert_eq!(obj["max_tokens"], 64); + + let back: SamplingParams = serde_json::from_value(value).unwrap(); + assert_eq!(back, params); + } } diff --git a/tests/entrypoints/scale_out/token_in_token_out/test_protocol.py b/tests/entrypoints/scale_out/token_in_token_out/test_protocol.py index 6a57e3b2b4..842dded1c7 100644 --- a/tests/entrypoints/scale_out/token_in_token_out/test_protocol.py +++ b/tests/entrypoints/scale_out/token_in_token_out/test_protocol.py @@ -132,6 +132,23 @@ def test_reasoning_parser_kwargs_rejects_unknown_keys(): GenerateRequest.model_validate(payload) +def test_rust_render_sampling_params_without_nulls_validate(): + """Rust render omits unset sampling fields; Python generate rejects nulls.""" + sampling_params = {"max_tokens": 64, "watermarking": True} + request = GenerateRequest.model_validate( + {"token_ids": [1, 2, 3], "sampling_params": sampling_params} + ) + assert request.sampling_params.max_tokens == 64 + + with pytest.raises(ValidationError, match="temperature"): + GenerateRequest.model_validate( + { + "token_ids": [1, 2, 3], + "sampling_params": {**sampling_params, "temperature": None}, + } + ) + + def test_output_mode_defaults_to_tokens(): assert GenerateRequest.model_validate(_base_payload()).output_mode == "tokens"