add visualization and warmup

app.py

@@ -358,7 +358,7 @@ class MedicalNERApp:
                     overflow: hidden;">
             <div style="background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
                         color: white; padding: 15px; text-align: center;">
-                <h3 style="margin: 0; font-size: 18px;">�� {model_name}</h3>
+                <h3 style="margin: 0; font-size: 18px;">{model_name}</h3>
                 <p style="margin: 5px 0 0 0; opacity: 0.9; font-size: 14px;">
                     Found {entity_count} medical entities
                 </p>
@@ -395,6 +395,19 @@ class MedicalNERApp:
                 f"{'...' if len(unique_texts) > 3 else ''}\n"
             )
 
+        # Add BIO tags information
+        summary_parts.append("\n🏷️ **BIO Tagging Info**\n")
+        summary_parts.append("The model uses BIO (Beginning-Inside-Outside) tagging scheme:\n")
+        summary_parts.append("• `B-LABEL`: Beginning of an entity\n")
+        summary_parts.append("• `I-LABEL`: Inside/continuation of an entity\n")
+        summary_parts.append("• `O`: Outside any entity (not shown in results)\n")
+
+        # Show example BIO tags for detected entity types
+        if entity_counts:
+            summary_parts.append("\nDetected entity types with their BIO tags:\n")
+            for label in sorted(entity_counts.keys()):
+                summary_parts.append(f"• `B-{label}`, `I-{label}`: {label} entities\n")
+
         return "\n".join(summary_parts)
 
 
@@ -402,6 +415,18 @@ class MedicalNERApp:
 print("🚀 Initializing Medical NER Application...")
 ner_app = MedicalNERApp()
 
+# Run a short warmup for each model here so it's not the first time
+print("🔥 Warming up models...")
+warmup_text = "The patient has diabetes and takes metformin."
+for model_name in MODELS.keys():
+    if ner_app.pipelines[model_name] is not None:
+        try:
+            print(f"Warming up {model_name}...")
+            _ = ner_app.predict_entities(warmup_text, model_name)
+            print(f"✅ {model_name} warmed up successfully")
+        except Exception as e:
+            print(f"⚠️ Warmup failed for {model_name}: {str(e)}")
+print("🎉 Model warmup complete!")
 
 def predict_wrapper(text: str, model_name: str):
     """Wrapper function for Gradio interface"""

requirements.txt

@@ -4,4 +4,5 @@ torch
 tokenizers
 numpy
 accelerate
-safetensors
+safetensors
+spacy