add app.py

app.py

@@ -0,0 +1,56 @@
+import gradio as gr
+import torch
+from transformers import AutoModel, AutoTokenizer
+
+
+model_path = 'CjangCjengh/NomBert-hn2qn-v0.1'
+device = 'cuda' if torch.cuda.is_available() else 'cpu'
+
+model = AutoModel.from_pretrained(model_path, torch_dtype='auto', trust_remote_code=True).eval().to(device)
+tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
+
+def parse_text(input_text):
+    with torch.inference_mode():
+        output_text, output_probs = model.parse_nom_text(tokenizer, [input_text])
+
+    html_content = '<div>'
+    for item in output_probs[0]:
+        char = item['char']
+        candidates = item['candidates']
+        html_content += f'<h3>{char}</h3>'
+        html_content += '<div style=\'display: flex; flex-wrap: wrap; gap: 10px;\'>'
+        
+        for candidate, prob in candidates:
+            prob_percent = prob * 100
+            html_content += f'''
+            <div style='margin-bottom: 15px; width: 170px;'>
+                <div style='margin-bottom: 5px;'>{candidate}: {prob_percent:.2f}%</div>
+                <div style='background-color: #f0f0f0; width: 100%; height: 15px; border-radius: 3px;'>
+                    <div style='background-color: #4caf50; width: {prob_percent}%; height: 100%; border-radius: 3px;'></div>
+                </div>
+            </div>
+            '''
+        html_content += '</div>'
+    html_content += '</div>'
+    
+    return output_text[0], html_content
+
+if __name__=='__main__':
+    with gr.Blocks(css='#viz {height: 500px; overflow-y: scroll;}') as app:
+        gr.Markdown('## NomBERT - Hán Nôm to Quốc Ngữ Converter')
+        with gr.Row():
+            with gr.Column(scale=1):
+                input_text = gr.Textbox(label='Input Hán Nôm Text', lines=5, placeholder='Enter Hán Nôm text here...')
+                parse_button = gr.Button('Parse')
+                output_text = gr.Textbox(label='Output Quốc Ngữ Text', lines=5, interactive=False)
+
+            with gr.Column(scale=2):
+                visualization = gr.HTML(label='Candidates Probabilities', elem_id='viz')
+
+        parse_button.click(
+            fn=parse_text,
+            inputs=input_text,
+            outputs=[output_text, visualization]
+        )
+
+    app.launch()

requirements.txt

@@ -0,0 +1 @@
+transformers==4.49.0