Update app.py

app.py

@@ -19,6 +19,7 @@ import fitz
 
 from transformers import (
     Qwen2_5_VLForConditionalGeneration,
+    AutoModelForVision2Seq,
     AutoModelForImageTextToText,
     AutoProcessor,
     TextIteratorStreamer,
@@ -59,10 +60,10 @@ model_t = AutoModelForImageTextToText.from_pretrained(
     MODEL_ID_T, trust_remote_code=True, torch_dtype=torch.float16
 ).to(device).eval()
 
-MODEL_ID_C = "nanonets/Nanonets-OCR-s"
+MODEL_ID_C = "HuggingFaceTB/SmolVLM-Instruct-250M"
 processor_c = AutoProcessor.from_pretrained(MODEL_ID_C, trust_remote_code=True)
-model_c = Qwen2_5_VLForConditionalGeneration.from_pretrained(
-    MODEL_ID_C, trust_remote_code=True, torch_dtype=torch.float16
+model_c = AutoModelForVision2Seq.from_pretrained(
+    MODEL_ID_C, trust_remote_code=True, torch_dtype=torch.float16, _attn_implementation="flash_attention_2"
 ).to(device).eval()
 
 MODEL_ID_G = "echo840/MonkeyOCR"
@@ -173,7 +174,7 @@ def process_document_stream(
         
     if model_name == "LFM2-VL-450M": processor, model = processor_m, model_m
     elif model_name == "LFM2-VL-1.6B": processor, model = processor_t, model_t
-    elif model_name == "Nanonets-OCR-s": processor, model = processor_c, model_c
+    elif model_name == "SmolVLM-Instruct-250M": processor, model = processor_c, model_c
     elif model_name == "MonkeyOCR-Recognition": processor, model = processor_g, model_g
     elif model_name == "olmOCR-7B-0725": processor, model = processor_i, model_i
     else:
@@ -232,7 +233,7 @@ def create_gradio_interface():
             # Left Column (Inputs)
             with gr.Column(scale=1):
                 model_choice = gr.Dropdown(
-                    choices=["LFM2-VL-1.6B", "LFM2-VL-450M", "olmOCR-7B-0725", "Nanonets-OCR-s", "Megalodon-OCR-Sync-0713"],
+                    choices=["LFM2-VL-1.6B", "LFM2-VL-450M", "SmolVLM-Instruct-250M", "olmOCR-7B-0725", "Megalodon-OCR-Sync-0713"],
                     label="Select Model", value="Nanonets-OCR-s"
                 )
                 prompt_input = gr.Textbox(label="Query Input", placeholder="✦︎ Enter your query")