Skip to content

fix: incorrect image token for Qwen2-VL #1

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
BugenZhao merged 2 commits into from
Jun 26, 2026
+7 −7
Merged

fix: incorrect image token for Qwen2-VL #1

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
625f2cd
correct qwen2-vl image_token
Isotr0py Jun 6, 2026
38065e2
Merge branch 'main' into fix-qwen2-vl
BugenZhao Jun 26, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 7 additions & 7 deletions src/registry/qwen_vl.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,10 +13,10 @@ pub(super) struct QwenVLVisionSpec;
impl QwenVLVisionSpec {
fn pad_token_id(metadata: &ModelMetadata) -> RegistryResult<TokenId> {
metadata
.config_u32(&["vision_token_id"])
.config_u32(&["image_token_id"])
.map(|v| v as TokenId)
.ok_or_else(|| ModelRegistryError::MissingConfigField {
field: "vision_token_id".to_string(),
field: "image_token_id".to_string(),
})
}
}
Expand All @@ -35,7 +35,7 @@ impl ModelProcessorSpec for QwenVLVisionSpec {
}

fn placeholder_token(&self, _metadata: &ModelMetadata) -> RegistryResult<String> {
Ok("<image>".to_string())
Ok("<|image_pad|>".to_string())
}

fn placeholder_token_id(&self, metadata: &ModelMetadata) -> RegistryResult<TokenId> {
Expand Down Expand Up @@ -104,7 +104,7 @@ mod tests {

#[test]
fn qwen_vision_uses_config_token_ids() {
let tokenizer = TestTokenizer::new(&[("<image>", 999)]);
let tokenizer = TestTokenizer::new(&[("<|image_pad|>", 151655)]);
let config = json!({
"model_type": "qwen2_vl",
"vision_start_token_id": 151652,
Expand All @@ -128,13 +128,13 @@ mod tests {
.unwrap();
// Only pad tokens — vision_start/vision_end are already in the chat template
assert_eq!(replacements[0].tokens.len(), 256);
assert_eq!(replacements[0].tokens[0], 151654); // pad (vision_token_id)
assert_eq!(*replacements[0].tokens.last().unwrap(), 151654);
assert_eq!(replacements[0].tokens[0], 151655); // pad (image_token_id)
assert_eq!(*replacements[0].tokens.last().unwrap(), 151655);
}

#[test]
fn qwen_vl_matches_alias_via_model_type() {
let tokenizer = TestTokenizer::new(&[("<image>", 999)]);
let tokenizer = TestTokenizer::new(&[("<|image_pad|>", 151655)]);
let config = json!({
"model_type": "qwen2_vl",
"vision_start_token_id": 151652,
Expand Down