Preserve image-aware sparse QSA for Qwen vision

src/engine/qwen_vision.rs

@@ -207,12 +207,27 @@ pub(super) struct Identity {
     pub(super) start: usize,
     pub(super) end: usize,
     pub(super) digest: [u8; 32],
+    /// Missing on checkpoints written before image-aware sparse QSA.
+    #[serde(default)]
+    pub(super) policy: u32,
+}
+
+pub(super) fn cache_policy() -> u32 {
+    u32::from(std::env::var("MTPLX_QWEN4_VISION_QSA").as_deref() != Ok("0"))
+}
+
+pub(super) fn compatible(images: &[Identity]) -> bool {
+    images.iter().all(|image| image.policy == cache_policy())
 }
 
 pub(super) fn same_prefix(a: &[Identity], b: &[Identity], tokens: usize) -> bool {
-    a.iter()
-        .filter(|i| i.start < tokens)
-        .eq(b.iter().filter(|i| i.start < tokens))
+    // Image pads do not identify partial image embeddings. MTPLX only restores
+    // before an image or after its complete span, never into the middle.
+    !b.iter()
+        .any(|image| image.start < tokens && tokens < image.end)
+        && a.iter()
+            .filter(|i| i.start < tokens)
+            .eq(b.iter().filter(|i| i.start < tokens))
 }
 
 #[test]
@@ -271,14 +286,19 @@ fn qwen_image_cache_keys_distinguish_pixels_and_preserve_prior_text() {
         start: 10,
         end: 14,
         digest: [1; 32],
+        policy: cache_policy(),
     }];
     let changed = vec![Identity {
         start: 10,
         end: 14,
         digest: [2; 32],
+        policy: cache_policy(),
     }];
     assert!(same_prefix(&old, &old, 20));
     assert!(!same_prefix(&old, &changed, 20));
     assert!(same_prefix(&old, &changed, 10));
     assert!(!same_prefix(&old, &[], 20));
+    assert!(!same_prefix(&old, &old, 11));
+    assert!(!same_prefix(&old, &old, 13));
+    assert!(same_prefix(&old, &old, 14));
 }