From 930f8dc0a1f2d19a0a61df3c0ab9b67dde36214e Mon Sep 17 00:00:00 2001
From: Reid <61492567+reidliu41@users.noreply.github.com>
Date: Tue, 30 Jun 2026 13:10:07 +0800
Subject: [PATCH] [Bugfix][Rust Frontend] Reject prompt_logprobs for streaming
 generate (#46839)

Signed-off-by: reidliu41 <reid201711@gmail.com>
---
 .../src/routes/inference/generate/validate.rs | 72 ++++++++++++++++---
 rust/src/server/src/routes/tests.rs           | 39 ++++++++++
 2 files changed, 103 insertions(+), 8 deletions(-)

diff --git a/rust/src/server/src/routes/inference/generate/validate.rs b/rust/src/server/src/routes/inference/generate/validate.rs
index 43347c60b5..3b925f3af2 100644
--- a/rust/src/server/src/routes/inference/generate/validate.rs
+++ b/rust/src/server/src/routes/inference/generate/validate.rs
@@ -34,14 +34,20 @@ pub(super) fn validate_request_compat(
         );
     }
 
-    if let Some(prompt_logprobs) = request.sampling_params.prompt_logprobs
-        && prompt_logprobs < 0
-        && prompt_logprobs != -1
-    {
-        bail_invalid_request!(
-            param = "sampling_params",
-            "`prompt_logprobs` must be a non-negative value or -1."
-        );
+    if let Some(prompt_logprobs) = request.sampling_params.prompt_logprobs {
+        if prompt_logprobs < 0 && prompt_logprobs != -1 {
+            bail_invalid_request!(
+                param = "sampling_params",
+                "`prompt_logprobs` must be a non-negative value or -1."
+            );
+        }
+
+        if request.stream {
+            bail_invalid_request!(
+                param = "sampling_params",
+                "`prompt_logprobs` are not available when `stream=true`."
+            );
+        }
     }
 
     Ok(())
@@ -97,4 +103,54 @@ mod tests {
         };
         assert!(validate_request_compat(&request, &served(&["Qwen/Qwen1.5-0.5B-Chat"])).is_err());
     }
+
+    #[test]
+    fn validate_request_compat_rejects_streaming_prompt_logprobs() {
+        let request: GenerateRequest = serde_json::from_value(json!({
+            "model": "Qwen/Qwen1.5-0.5B-Chat",
+            "token_ids": [11, 22],
+            "stream": true,
+            "sampling_params": {
+                "prompt_logprobs": 0
+            }
+        }))
+        .expect("parse request");
+        assert!(validate_request_compat(&request, &served(&["Qwen/Qwen1.5-0.5B-Chat"])).is_err());
+
+        let request: GenerateRequest = serde_json::from_value(json!({
+            "model": "Qwen/Qwen1.5-0.5B-Chat",
+            "token_ids": [11, 22],
+            "stream": true,
+            "sampling_params": {
+                "prompt_logprobs": 1
+            }
+        }))
+        .expect("parse request");
+        assert!(validate_request_compat(&request, &served(&["Qwen/Qwen1.5-0.5B-Chat"])).is_err());
+
+        let request: GenerateRequest = serde_json::from_value(json!({
+            "model": "Qwen/Qwen1.5-0.5B-Chat",
+            "token_ids": [11, 22],
+            "stream": true,
+            "sampling_params": {
+                "prompt_logprobs": -1
+            }
+        }))
+        .expect("parse request");
+        assert!(validate_request_compat(&request, &served(&["Qwen/Qwen1.5-0.5B-Chat"])).is_err());
+    }
+
+    #[test]
+    fn validate_request_compat_accepts_non_stream_prompt_logprobs() {
+        let request: GenerateRequest = serde_json::from_value(json!({
+            "model": "Qwen/Qwen1.5-0.5B-Chat",
+            "token_ids": [11, 22],
+            "stream": false,
+            "sampling_params": {
+                "prompt_logprobs": 1
+            }
+        }))
+        .expect("parse request");
+        assert!(validate_request_compat(&request, &served(&["Qwen/Qwen1.5-0.5B-Chat"])).is_ok());
+    }
 }
diff --git a/rust/src/server/src/routes/tests.rs b/rust/src/server/src/routes/tests.rs
index 0c3450caf1..d7e99e8fa9 100644
--- a/rust/src/server/src/routes/tests.rs
+++ b/rust/src/server/src/routes/tests.rs
@@ -4129,6 +4129,45 @@ async fn raw_generate_rejects_empty_token_ids() {
     assert_eq!(json["error"]["param"], "token_ids");
 }
 
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+#[serial]
+async fn raw_generate_rejects_streaming_prompt_logprobs() {
+    let mut app = test_app().await;
+
+    for prompt_logprobs in [0, 1] {
+        let response = app
+            .call(
+                Request::builder()
+                    .method("POST")
+                    .uri("/inference/v1/generate")
+                    .header("content-type", "application/json")
+                    .body(Body::from(
+                        json!({
+                            "model": "Qwen/Qwen1.5-0.5B-Chat",
+                            "token_ids": [11, 22],
+                            "stream": true,
+                            "sampling_params": {
+                                "prompt_logprobs": prompt_logprobs
+                            }
+                        })
+                        .to_string(),
+                    ))
+                    .expect("build request"),
+            )
+            .await
+            .expect("call app");
+
+        assert_eq!(response.status(), StatusCode::BAD_REQUEST);
+        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");
+        assert_eq!(json["error"]["param"], "sampling_params");
+        assert_eq!(
+            json["error"]["message"],
+            "`prompt_logprobs` are not available when `stream=true`."
+        );
+    }
+}
+
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 #[serial]
 async fn raw_generate_rejects_wrong_model() {