Update app.py

app.py

@@ -27,6 +27,7 @@ from transformers import (
     AutoModel,
     AutoProcessor,
     TextIteratorStreamer,
+    AutoTokenizer,
 )
 
 from transformers.image_utils import load_image
@@ -103,6 +104,19 @@ model_z = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     MODEL_ID_Z, trust_remote_code=True, torch_dtype=torch.float16
 ).to(device).eval()
 
+# --- Moondream2 Model Loading ---
+MODEL_ID_MD = "vikhyatk/moondream2"
+REVISION_MD = "2025-06-21"
+moondream = AutoModelForCausalLM.from_pretrained(
+  MODEL_ID_MD,
+  revision=REVISION_MD,
+  trust_remote_code=True,
+  torch_dtype=torch.float16,
+  device_map={"": "cuda"},
+)
+tokenizer_md = AutoTokenizer.from_pretrained(MODEL_ID_MD, revision=REVISION_MD)
+
+
 # --- PDF Generation and Preview Utility Function ---
 def generate_and_preview_pdf(image: Image.Image, text_content: str, font_size: int, line_spacing: float, alignment: str, image_size: str):
     """
@@ -192,7 +206,17 @@ def process_document_stream(
     if not prompt_input or not prompt_input.strip():
         yield "Please enter a prompt.", ""
         return
-        
+
+    if model_name == "Moondream2(vision)":
+        image_embeds = moondream.encode_image(image)
+        answer = moondream.answer_question(
+            image_embeds=image_embeds,
+            question=prompt_input,
+            tokenizer=tokenizer_md
+        )
+        yield answer, answer
+        return
+
     if model_name == "LFM2-VL-450M(fast)": processor, model = processor_m, model_m
     elif model_name == "LFM2-VL-1.6B(fast)": processor, model = processor_t, model_t
     elif model_name == "ShotVL-3B(cinematic)": processor, model = processor_z, model_z
@@ -257,9 +281,9 @@ def create_gradio_interface():
             # Left Column (Inputs)
             with gr.Column(scale=1):
                 model_choice = gr.Dropdown(
-                    choices=["LFM2-VL-1.6B(fast)", "LFM2-VL-450M(fast)", "SmolVLM-Instruct-250M(smol)", "ShotVL-3B(cinematic)", "Megalodon-OCR-Sync-0713(ocr)", 
+                    choices=["LFM2-VL-450M(fast)", "LFM2-VL-1.6B(fast)", "SmolVLM-Instruct-250M(smol)", "Moondream2(vision)", "ShotVL-3B(cinematic)", "Megalodon-OCR-Sync-0713(ocr)", 
                              "VLAA-Thinker-Qwen2VL-2B(reason)", "MonkeyOCR-pro-1.2B(ocr)", "Nanonets-OCR-s(ocr)"],
-                    label="Select Model", value="LFM2-VL-450M(fast)"
+                    label="Select Model", value= "LFM2-VL-450M(fast)"
                 )
                 prompt_input = gr.Textbox(label="Query Input", placeholder="✦︎ Enter your query", value="Describe the image!")
                 image_input = gr.Image(label="Upload Image", type="pil", sources=['upload'])