Update app.py

app.py

@@ -19,10 +19,13 @@ import fitz
 
 from transformers import (
     Qwen2_5_VLForConditionalGeneration,
+    AutoModelForImageTextToText,
     AutoProcessor,
     TextIteratorStreamer,
 )
 
+from transformers.image_utils import load_image
+
 from reportlab.lib.pagesizes import A4
 from reportlab.lib.styles import getSampleStyleSheet
 from reportlab.platypus import SimpleDocTemplate, Image as RLImage, Paragraph, Spacer
@@ -44,15 +47,15 @@ if torch.cuda.is_available():
 print("Using device:", device)
 
 # --- Model Loading ---
-MODEL_ID_M = "prithivMLmods/Camel-Doc-OCR-080125"
+MODEL_ID_M = "LiquidAI/LFM2-VL-450M"
 processor_m = AutoProcessor.from_pretrained(MODEL_ID_M, trust_remote_code=True)
-model_m = Qwen2_5_VLForConditionalGeneration.from_pretrained(
+model_m = AutoModelForImageTextToText.from_pretrained(
     MODEL_ID_M, trust_remote_code=True, torch_dtype=torch.float16
 ).to(device).eval()
 
-MODEL_ID_T = "prithivMLmods/Megalodon-OCR-Sync-0713"
+MODEL_ID_T = "LiquidAI/LFM2-VL-1.6B"
 processor_t = AutoProcessor.from_pretrained(MODEL_ID_T, trust_remote_code=True)
-model_t = Qwen2_5_VLForConditionalGeneration.from_pretrained(
+model_t = AutoModelForImageTextToText.from_pretrained(
     MODEL_ID_T, trust_remote_code=True, torch_dtype=torch.float16
 ).to(device).eval()
 
@@ -168,8 +171,8 @@ def process_document_stream(
         yield "Please enter a prompt.", ""
         return
         
-    if model_name == "Camel-Doc-OCR-080125": processor, model = processor_m, model_m
-    elif model_name == "Megalodon-OCR-Sync-0713": processor, model = processor_t, model_t
+    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 == "MonkeyOCR-Recognition": processor, model = processor_g, model_g
     elif model_name == "olmOCR-7B-0725": processor, model = processor_i, model_i
@@ -229,7 +232,7 @@ def create_gradio_interface():
             # Left Column (Inputs)
             with gr.Column(scale=1):
                 model_choice = gr.Dropdown(
-                    choices=["Camel-Doc-OCR-080125", "MonkeyOCR-Recognition", "olmOCR-7B-0725", "Nanonets-OCR-s", "Megalodon-OCR-Sync-0713"],
+                    choices=["LFM2-VL-1.6B", "LFM2-VL-450M", "olmOCR-7B-0725", "Nanonets-OCR-s", "Megalodon-OCR-Sync-0713"],
                     label="Select Model", value="Nanonets-OCR-s"
                 )
                 prompt_input = gr.Textbox(label="Query Input", placeholder="✦︎ Enter your query")