update app (#12)

- update app (69e5cb42fe5bfb1aa419bd1dabf4058cb10d58b8)

app.py

@@ -150,6 +150,16 @@ model_o = AutoModelForVision2Seq.from_pretrained(
     MODEL_ID_O, trust_remote_code=True, torch_dtype=torch.float16, _attn_implementation="flash_attention_2"
 ).to(device).eval()
 
+# --- NEW MODEL: SmolVLM2-500M-Video-Instruct ---
+MODEL_ID_SV = "HuggingFaceTB/SmolVLM2-500M-Video-Instruct"
+processor_sv = AutoProcessor.from_pretrained(MODEL_ID_SV, trust_remote_code=True)
+model_sv = AutoModelForImageTextToText.from_pretrained(
+    MODEL_ID_SV,
+    trust_remote_code=True,
+    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):
     """
@@ -250,6 +260,7 @@ def process_document_stream(
         yield answer, answer
         return
 
+    # Model and processor selection
     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
@@ -263,6 +274,7 @@ def process_document_stream(
     elif model_name == "TBAC-VLR1-3B(open-r1)": processor, model = processor_g, model_g
     elif model_name == "OCRFlux-3B(ocr)": processor, model = processor_v, model_v
     elif model_name == "SmolVLM-500M-Instruct(smol)": processor, model = processor_o, model_o
+    elif model_name == "SmolVLM2-500M-Video-Instruct(video)": processor, model = processor_sv, model_sv
     else:
         yield "Invalid model selected.", ""
         return
@@ -321,7 +333,7 @@ def create_gradio_interface():
                 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)", 
                              "VLAA-Thinker-Qwen2VL-2B(reason)", "MonkeyOCR-pro-1.2B(ocr)", "Qwen2.5-VL-3B-Abliterated-Caption-it(caption)", "Nanonets-OCR-s(ocr)",
-                             "LMM-R1-MGT-PerceReason(reason)", "OCRFlux-3B(ocr)", "TBAC-VLR1-3B(open-r1)", "SmolVLM-500M-Instruct(smol)"],
+                             "LMM-R1-MGT-PerceReason(reason)", "OCRFlux-3B(ocr)", "TBAC-VLR1-3B(open-r1)", "SmolVLM-500M-Instruct(smol)", "SmolVLM2-500M-Video-Instruct(video)"],
                     label="Select Model", value= "LFM2-VL-450M(fast)"
                 )
                 prompt_input = gr.Textbox(label="Query Input", placeholder="✦︎ Enter your query", value="Describe the image!")