[Model]: support KeyeVL-1_5-8B (#23838)

Signed-off-by: wangruitao <wangruitao@kuaishou.com>
Co-authored-by: wangruitao <wangruitao@kuaishou.com>

examples/offline_inference/vision_language.py

@@ -683,6 +683,37 @@ def run_keye_vl(questions: list[str], modality: str) -> ModelRequestData:
     )
 
 
+# Keye-VL-1.5
+def run_keye_vl1_5(questions: list[str], modality: str) -> ModelRequestData:
+    model_name = "Kwai-Keye/Keye-VL-1.5-8B"
+
+    engine_args = EngineArgs(
+        model=model_name,
+        max_model_len=8192,
+        trust_remote_code=True,
+        limit_mm_per_prompt={modality: 1},
+    )
+
+    if modality == "image":
+        placeholder = "<|image_pad|>"
+    elif modality == "video":
+        placeholder = "<|video_pad|>"
+
+    prompts = [
+        (
+            f"<|im_start|>user\n<|vision_start|>{placeholder}<|vision_end|>"
+            f"{question}<|im_end|>\n"
+            "<|im_start|>assistant\n"
+        )
+        for question in questions
+    ]
+
+    return ModelRequestData(
+        engine_args=engine_args,
+        prompts=prompts,
+    )
+
+
 # Kimi-VL
 def run_kimi_vl(questions: list[str], modality: str) -> ModelRequestData:
     assert modality == "image"
@@ -1648,6 +1679,7 @@ model_example_map = {
     "interns1": run_interns1,
     "internvl_chat": run_internvl,
     "keye_vl": run_keye_vl,
+    "keye_vl1_5": run_keye_vl1_5,
     "kimi_vl": run_kimi_vl,
     "llama4": run_llama4,
     "llava": run_llava,

examples/offline_inference/vision_language_multi_image.py

@@ -542,6 +542,43 @@ def load_keye_vl(question: str, image_urls: list[str]) -> ModelRequestData:
     )
 
 
+def load_keye_vl1_5(question: str, image_urls: list[str]) -> ModelRequestData:
+    model_name = "Kwai-Keye/Keye-VL-1_5-8B"
+
+    engine_args = EngineArgs(
+        model=model_name,
+        trust_remote_code=True,
+        max_model_len=8192,
+        max_num_seqs=5,
+        limit_mm_per_prompt={"image": len(image_urls)},
+    )
+
+    placeholders = [{"type": "image", "image": url} for url in image_urls]
+    messages = [
+        {
+            "role": "user",
+            "content": [
+                *placeholders,
+                {"type": "text", "text": question},
+            ],
+        },
+    ]
+
+    processor = AutoProcessor.from_pretrained(model_name, trust_remote_code=True)
+
+    prompt = processor.apply_chat_template(
+        messages, tokenize=False, add_generation_prompt=True
+    )
+
+    image_data = [fetch_image(url) for url in image_urls]
+
+    return ModelRequestData(
+        engine_args=engine_args,
+        prompt=prompt,
+        image_data=image_data,
+    )
+
+
 def load_kimi_vl(question: str, image_urls: list[str]) -> ModelRequestData:
     model_name = "moonshotai/Kimi-VL-A3B-Instruct"
 
@@ -1209,6 +1246,7 @@ model_example_map = {
     "interns1": load_interns1,
     "internvl_chat": load_internvl,
     "keye_vl": load_keye_vl,
+    "keye_vl1_5": load_keye_vl1_5,
     "kimi_vl": load_kimi_vl,
     "llama4": load_llama4,
     "llava": load_llava,