first

app.py

@@ -1,7 +1,103 @@
 import gradio as gr
+import torch
+from transformers import AutoModelForVision2Seq, AutoProcessor, BitsAndBytesConfig
+from PIL import Image
 
-def greet(name):
-    return "Hello " + name + "!!"
+# Global variables for model and processor
+model = None
+processor = None
 
-demo = gr.Interface(fn=greet, inputs="text", outputs="text")
-demo.launch()
+def load_model_and_processor():
+    global model, processor
+    
+    model_path = "Aekanun/thai-handwriting-llm"
+    base_model_path = "meta-llama/Llama-3.2-11B-Vision-Instruct"
+    
+    # BitsAndBytes config for 4-bit quantization
+    bnb_config = BitsAndBytesConfig(
+        load_in_4bit=True,
+        bnb_4bit_use_double_quant=True,
+        bnb_4bit_quant_type="nf4",
+        bnb_4bit_compute_dtype=torch.bfloat16
+    )
+    
+    try:
+        # Load processor from base model
+        processor = AutoProcessor.from_pretrained(base_model_path)
+        
+        # Load fine-tuned model
+        model = AutoModelForVision2Seq.from_pretrained(
+            model_path,
+            device_map="auto",
+            torch_dtype=torch.bfloat16,
+            quantization_config=bnb_config
+        )
+        return True
+    except Exception as e:
+        print(f"Error loading model: {str(e)}")
+        return False
+
+def process_handwriting(image):
+    global model, processor
+    
+    if image is None:
+        return "กรุณาอัพโหลดรูปภาพ"
+    
+    try:
+        # Ensure image is in PIL format
+        if not isinstance(image, Image.Image):
+            image = Image.fromarray(image)
+        
+        # Prepare prompt and messages
+        prompt = """Transcribe the Thai handwritten text from the provided image.
+Only return the transcription in Thai language."""
+        messages = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": prompt},
+                    {"type": "image", "image": image}
+                ],
+            }
+        ]
+        
+        # Process input
+        text = processor.apply_chat_template(messages, tokenize=False)
+        inputs = processor(text=text, images=image, return_tensors="pt")
+        inputs = {k: v.to(model.device) for k, v in inputs.items()}
+        
+        # Generate output
+        with torch.no_grad():
+            outputs = model.generate(
+                **inputs,
+                max_new_tokens=256,
+                do_sample=False,
+                pad_token_id=processor.tokenizer.pad_token_id
+            )
+        
+        # Decode output
+        transcription = processor.decode(outputs[0], skip_special_tokens=True)
+        return transcription
+    
+    except Exception as e:
+        return f"เกิดข้อผิดพลาด: {str(e)}"
+
+# Load model when starting
+print("กำลังโหลดโมเดล...")
+model_loaded = load_model_and_processor()
+
+if model_loaded:
+    # Create Gradio interface
+    demo = gr.Interface(
+        fn=process_handwriting,
+        inputs=gr.Image(type="pil", label="อัพโหลดรูปลายมือเขียนภาษาไทย"),
+        outputs=gr.Textbox(label="ข้อความที่แปลงได้"),
+        title="Thai Handwriting to Text ด้วย LLaMA Vision",
+        description="อัพโหลดรูปภาพลายมือเขียนภาษาไทยเพื่อแปลงเป็นข้อความ โดยใช้โมเดล LLaMA Vision ที่ fine-tune มาสำหรับภาษาไทย",
+        examples=[["example1.jpg"], ["example2.jpg"]]
+    )
+    
+    if __name__ == "__main__":
+        demo.launch(share=True)
+else:
+    print("ไม่สามารถโหลดโมเดลได้ กรุณาตรวจสอบ log")

requirements.txt

@@ -0,0 +1,6 @@
+gradio==4.13.0
+torch>=2.0.0
+transformers>=4.34.0
+Pillow>=9.0.0
+bitsandbytes>=0.41.1
+accelerate>=0.24.1