Update app.py

app.py

@@ -116,6 +116,16 @@ moondream = AutoModelForCausalLM.from_pretrained(
 )
 tokenizer_md = AutoTokenizer.from_pretrained(MODEL_ID_MD, revision=REVISION_MD)
 
+# --- SmolVLM2 Model Loading ---
+MODEL_ID_S2 = "HuggingFaceTB/SmolVLM2-2.2B-Instruct"
+processor_s2 = AutoProcessor.from_pretrained(MODEL_ID_S2, trust_remote_code=True)
+model_s2 = AutoModelForImageTextToText.from_pretrained(
+    MODEL_ID_S2, 
+    trust_remote_code=True,
+    _attn_implementation="flash_attention_2",
+    torch_dtype=torch.float16
+).to(device).eval()
+
 
 # --- 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):
@@ -217,7 +227,8 @@ def process_document_stream(
         yield answer, answer
         return
 
-    if model_name == "LFM2-VL-450M(fast)": processor, model = processor_m, model_m
+    if model_name == "SmolVLM2-2.2B-Instruct(smol)": processor, model = processor_s2, model_s2
+    elif 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
     elif model_name == "SmolVLM-Instruct-250M(smol)": processor, model = processor_c, model_c
@@ -281,9 +292,9 @@ def create_gradio_interface():
             # Left Column (Inputs)
             with gr.Column(scale=1):
                 model_choice = gr.Dropdown(
-                    choices=["LFM2-VL-450M(fast)", "LFM2-VL-1.6B(fast)", "SmolVLM-Instruct-250M(smol)", "Moondream2(vision)", "ShotVL-3B(cinematic)", "Megalodon-OCR-Sync-0713(ocr)", 
+                    choices=["SmolVLM2-2.2B-Instruct(smol)", "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= "SmolVLM2-2.2B-Instruct(smol)"
                 )
                 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'])