Update Gradio interface with configurable parameters and error handling

app.py

@@ -5,58 +5,52 @@ from transformers import T5Tokenizer, T5ForConditionalGeneration
 model = T5ForConditionalGeneration.from_pretrained('t5-small')
 tokenizer = T5Tokenizer.from_pretrained('t5-small')
 
-def generate_clinical_report(input_text):
+def generate_clinical_report(input_text, max_length=256, num_beams=4, no_repeat_ngram_size=3, length_penalty=2.0, early_stopping=True):
     """
-    Generate a clinical report from the input text using the fine-tuned T5 model.
+    Generate a clinical report from the input text using the T5 model with configurable parameters.
     """
-    # Prepare input text
-    input_ids = tokenizer.encode("summarize: " + input_text, return_tensors="pt", max_length=512, truncation=True)
-    
-    # Generate report
-    outputs = model.generate(
-        input_ids,
-        max_length=256,
-        num_beams=4,
-        no_repeat_ngram_size=3,
-        length_penalty=2.0,
-        early_stopping=True,
-        bad_words_ids=[[tokenizer.encode(word, add_special_tokens=False)[0]] 
-                      for word in ['http', 'www', '.com', '.org']]
-    )
-    
-    # Decode and return the generated report
-    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+    try:
+        # Prepare input text
+        input_ids = tokenizer.encode("summarize: " + input_text, return_tensors="pt", max_length=512, truncation=True)
+        
+        # Generate report with provided parameters
+        outputs = model.generate(
+            input_ids,
+            max_length=max_length,
+            num_beams=num_beams,
+            no_repeat_ngram_size=no_repeat_ngram_size,
+            length_penalty=length_penalty,
+            early_stopping=early_stopping,
+            bad_words_ids=[[tokenizer.encode(word, add_special_tokens=False)[0]] 
+                          for word in ['http', 'www', '.com', '.org']]
+        )
+        
+        # Decode and return the generated report
+        return tokenizer.decode(outputs[0], skip_special_tokens=True)
+    except Exception as e:
+        print(f"Error generating report: {str(e)}")
+        return f"Error: {str(e)}"
 
-# Create Gradio interface with queue
+# Create Gradio interface with API configuration
 demo = gr.Interface(
     fn=generate_clinical_report,
     inputs=[
         gr.Textbox(
             lines=8,
             placeholder="Enter clinical notes here...",
-            label="Clinical Notes",
-            elem_id="input-box"
-        )
-    ],
-    outputs=[
-        gr.Textbox(
-            lines=8,
-            label="Generated Clinical Report",
-            elem_id="output-box"
-        )
+            label="Clinical Notes"
+        ),
+        gr.Slider(minimum=50, maximum=500, value=256, step=1, label="Max Length"),
+        gr.Slider(minimum=1, maximum=8, value=4, step=1, label="Num Beams"),
+        gr.Slider(minimum=1, maximum=5, value=3, step=1, label="No Repeat Ngram Size"),
+        gr.Slider(minimum=0.1, maximum=5.0, value=2.0, step=0.1, label="Length Penalty"),
+        gr.Checkbox(value=True, label="Early Stopping")
     ],
+    outputs=gr.Textbox(lines=8, label="Generated Clinical Report"),
     title="Clinical Report Generator",
     description="Generate professional clinical reports from clinical notes using a T5 model.",
-    examples=[
-        ["Patient presented with severe abdominal pain in the lower right quadrant. Temperature 38.5°C, BP 130/85."],
-        ["Follow-up visit for diabetes management. Blood sugar levels have been stable with current medication regimen."]
-    ],
-    theme=gr.themes.Soft(),
-    css="""
-        #input-box { background-color: #f6f6f6; }
-        #output-box { background-color: #f0f7ff; }
-    """,
-    flagging_mode="never"
+    allow_flagging="never",
+    analytics_enabled=False
 )
 
 # Launch the app with optimized configuration for Hugging Face Spaces