[BugFix] Add BAGEL single-stage diffusion config and fix multiple <im_start><im_end> bug (#2381)

Signed-off-by: princepride <wangzhipeng628@gmail.com>

examples/offline_inference/bagel/end2end.py

@@ -101,7 +101,7 @@ def main():
 
     if not prompts:
         # Default prompt for text2img test if none provided
-        prompts = ["<|im_start|>A cute cat<|im_end|>"]
+        prompts = ["A cute cat"]
         print(f"[Info] No prompts provided, using default: {prompts}")
     omni_outputs = []
 

examples/online_serving/bagel/openai_chat_client.py

@@ -125,7 +125,7 @@ def generate_image(
 
 def main():
     parser = argparse.ArgumentParser(description="Bagel multimodal chat client")
-    parser.add_argument("--prompt", "-p", default="<|im_start|>A cute cat<|im_end|>", help="Text prompt")
+    parser.add_argument("--prompt", "-p", default="A cute cat", help="Text prompt")
     parser.add_argument("--output", "-o", default="bagel_output.png", help="Output file (for image results)")
     parser.add_argument("--server", "-s", default="http://localhost:8091", help="Server URL")
 

vllm_omni/model_executor/stage_configs/bagel_single_stage.yaml

@@ -0,0 +1,32 @@
+# Stage 0: Thinker (multimodal understanding + text generation)
+
+stage_args:
+
+  - stage_id: 0
+    stage_type: diffusion
+    runtime:
+      devices: "0"
+    engine_args:
+      model_stage: dit
+      max_num_seqs: 1
+      gpu_memory_utilization: 0.45
+      enforce_eager: true
+      trust_remote_code: true
+      engine_output_type: image
+      distributed_executor_backend: "mp"
+      enable_prefix_caching: false
+      max_num_batched_tokens: 32768
+      tensor_parallel_size: 1
+
+    final_output: true
+    final_output_type: image
+    is_comprehension: false
+    default_sampling_params:
+      seed: 52
+
+# Runtime edges
+runtime:
+  enabled: true
+  defaults:
+    window_size: -1
+    max_inflight: 1