updates

app.py

@@ -2,7 +2,7 @@ import gradio as gr
 from llama_cpp import Llama
 from huggingface_hub import hf_hub_download
 
-# Function to load a user-specified model from Hugging Face
+# Load a user-specified model
 def load_user_model(repo_id, model_file):
     print(f"Downloading model {model_file} from repository {repo_id}...")
     local_path = hf_hub_download(repo_id=repo_id, filename=model_file)
@@ -14,9 +14,12 @@ def generate_response(model, prompt):
     response = model(prompt, max_tokens=256, temperature=0.7)
     return response["choices"][0]["text"]
 
-# Evaluate responses generated by two models using the LoRA evaluation model
-def evaluate_responses(prompt, repo_a, model_a, repo_b, model_b, criteria_list):
-    # Load user-specified models
+# Evaluate responses using the LoRA evaluation model
+def evaluate_responses(prompt, repo_a, model_a, repo_b, model_b, evaluation_criteria):
+    if len(evaluation_criteria) > 3:
+        return "Error: Please select up to 3 evaluation criteria only."
+
+    # Load models
     model_a_instance = load_user_model(repo_a, model_a)
     model_b_instance = load_user_model(repo_b, model_b)
     
@@ -24,10 +27,12 @@ def evaluate_responses(prompt, repo_a, model_a, repo_b, model_b, criteria_list):
     response_a = generate_response(model_a_instance, prompt)
     response_b = generate_response(model_b_instance, prompt)
     
+    # Display generated responses
     print(f"Response A: {response_a}")
     print(f"Response B: {response_b}")
     
-    # Format the evaluation prompt for the LoRA model
+    # Format the evaluation prompt
+    criteria_list = ", ".join(evaluation_criteria)
     evaluation_prompt = f"""
 Prompt: {prompt}
 
@@ -45,15 +50,16 @@ Please evaluate the responses based on the selected criteria. For each criterion
         temperature=0.5
     )
     evaluation_results = evaluation_response["choices"][0]["text"]
-
+    
+    # Combine results for display
     final_output = f"""
-    Response A:\n{response_a}\n\n
-    Response B:\n{response_b}\n\n
-    Evaluation Results:\n{evaluation_results}
-    """
+Response A:\n{response_a}\n\n
+Response B:\n{response_b}\n\n
+Evaluation Results:\n{evaluation_results}
+"""
     return final_output
 
-# Load the base LoRA evaluation model
+# Load the LoRA evaluation model
 def load_lora_model():
     repo_id = "KolumbusLindh/LoRA-4100"
     model_file = "unsloth.F16.gguf"
@@ -69,41 +75,35 @@ print("LoRA evaluation model loaded successfully!")
 with gr.Blocks(title="LLM as a Judge") as demo:
     gr.Markdown("## LLM as a Judge 🧐")
 
-    # Inputs for Model A repository and file
-    repo_a_input = gr.Textbox(label="Model A Repository (e.g., KolumbusLindh/LoRA-4100)", placeholder="Enter the Hugging Face repo name for Model A...")
-    model_a_input = gr.Textbox(label="Model A File Name (e.g., unsloth.F16.gguf)", placeholder="Enter the model filename for Model A...")
-
-    # Inputs for Model B repository and file
-    repo_b_input = gr.Textbox(label="Model B Repository (e.g., KolumbusLindh/LoRA-4100)", placeholder="Enter the Hugging Face repo name for Model B...")
-    model_b_input = gr.Textbox(label="Model B File Name (e.g., unsloth.F16.gguf)", placeholder="Enter the model filename for Model B...")
+    # Model inputs
+    repo_a_input = gr.Textbox(label="Model A Repository", placeholder="Enter the Hugging Face repo name for Model A...")
+    model_a_input = gr.Textbox(label="Model A File Name", placeholder="Enter the model filename for Model A...")
+    repo_b_input = gr.Textbox(label="Model B Repository", placeholder="Enter the Hugging Face repo name for Model B...")
+    model_b_input = gr.Textbox(label="Model B File Name", placeholder="Enter the model filename for Model B...")
 
-    # Input for prompt and evaluation criteria
+    # Prompt and criteria inputs
     prompt_input = gr.Textbox(label="Enter Prompt", placeholder="Enter the prompt here...", lines=3)
     criteria_dropdown = gr.CheckboxGroup(
         label="Select Up to 3 Evaluation Criteria",
-        choices=["Clarity", "Completeness", "Accuracy", "Relevance", "User-Friendliness", "Depth", "Creativity"],
-        value=["Clarity"],
-        max_choices=3
+        choices=["Clarity", "Completeness", "Accuracy", "Relevance", "User-Friendliness", "Depth", "Creativity"]
     )
 
-    # Button to evaluate responses
+    # Button and outputs
     evaluate_button = gr.Button("Evaluate Models")
-    
-    # Output for evaluation results
     evaluation_output = gr.Textbox(
         label="Evaluation Results",
         placeholder="The evaluation results will appear here...",
-        lines=10,
+        lines=20,
         interactive=False
     )
 
-    # Link the evaluation function to the button
+    # Link evaluation function
     evaluate_button.click(
         fn=evaluate_responses,
         inputs=[prompt_input, repo_a_input, model_a_input, repo_b_input, model_b_input, criteria_dropdown],
         outputs=[evaluation_output]
     )
 
-# Launch the Gradio app
+# Launch app
 if __name__ == "__main__":
-    demo.launch()  # Add share=True to create a public link
+    demo.launch()