UI level changes

app.py

@@ -11,7 +11,6 @@ from generationhelper import GenerateAnswer
 from formatresultshelper import FormatAndScores
 from calculatescores import CalculateScores
 
-from model import generate_response
 from huggingface_hub import login
 from huggingface_hub import whoami
 from huggingface_hub import dataset_info
@@ -33,50 +32,80 @@ login(hf_token)
 rag_extracted_data = ExtractRagBenchData()
 print(rag_extracted_data.head(5))
 
-#invoke create milvus db function
-try:
-    db_collection = CreateMilvusDbSchema()
-except Exception as e:
-    print(f"Error creating Milvus DB schema: {e}")
-
-#insert embdeding to milvus db
 """
 EmbedAllDocumentsAndInsert(QUERY_EMBEDDING_MODEL, rag_extracted_data, db_collection, window_size=WINDOW_SIZE, overlap=OVERLAP)
 """  
-query = "what would the net revenue have been in 2015 if there wasn't a stipulated settlement from the business combination in october 2015?"
 
-results_for_top10_chunks = SearchTopKDocuments(db_collection, query, QUERY_EMBEDDING_MODEL, top_k=RETRIVE_TOP_K_SIZE)
+def EvaluateRAGModel(query, evaluation_model):
+    #invoke create milvus db function
+    try:
+        db_collection = CreateMilvusDbSchema()
+    except Exception as e:
+        print(f"Error creating Milvus DB schema: {e}")
+
+    #insert embdeding to milvus db
+
+    #query = "what would the net revenue have been in 2015 if there wasn't a stipulated settlement from the business combination in october 2015?"
+
+    results_for_top10_chunks = SearchTopKDocuments(db_collection, query, QUERY_EMBEDDING_MODEL, top_k=RETRIVE_TOP_K_SIZE)
+
+    reranked_results = FineTuneAndRerankSearchResults(results_for_top10_chunks, rag_extracted_data, query, RERANKING_MODEL)
+
+    answer = GenerateAnswer(query, reranked_results.head(3), PROMPT_MODEL)
+
+    completion_result,relevant_sentence_keys,all_utilized_sentence_keys,support_keys,support_level = FormatAndScores(query, reranked_results.head(1), answer, EVAL_MODEL)
+
+
+    print(relevant_sentence_keys)
+    print(all_utilized_sentence_keys)
+    print(support_keys)
+    print(support_level)
+    print(completion_result)
 
-reranked_results = FineTuneAndRerankSearchResults(results_for_top10_chunks, rag_extracted_data, query, RERANKING_MODEL)
+    document_id = reranked_results.head(1)['doc_id'].values[0]
+    extarcted_row_for_given_id = rag_extracted_data[rag_extracted_data["id"]==document_id]
 
-answer = GenerateAnswer(query, reranked_results.head(3), PROMPT_MODEL)
+    rmsecontextrel, rmsecontextutil, aucscore = CalculateScores(relevant_sentence_keys,all_utilized_sentence_keys,support_keys,support_level,extarcted_row_for_given_id)
 
-completion_result,relevant_sentence_keys,all_utilized_sentence_keys,support_keys,support_level = FormatAndScores(query, reranked_results.head(1), answer, EVAL_MODEL)
+    print(rmsecontextrel)
+    print(rmsecontextutil)
+    print(aucscore)
 
 
-print(relevant_sentence_keys)
-print(all_utilized_sentence_keys)
-print(support_keys)
-print(support_level)
-print(completion_result)
+# Create Gradio UI
+with gr.Blocks() as iface:
+    gr.Markdown("## Capstone Project Group 10 - Model Evaluation")
 
-document_id = reranked_results.head(1)['doc_id'].values[0]
-extarcted_row_for_given_id = rag_extracted_data[rag_extracted_data["id"]==document_id]
+    with gr.Row():
+        question_input = gr.Textbox(label="Enter your Question", lines=2)
+        dropdown_input = gr.Dropdown(
+            ["LLaMA 3.3", "Mistral &B", "Model C"], 
+            value="LLaMA 3.3", 
+            label="Select a Model"
+        )
 
-score1, score2, score3 = CalculateScores(relevant_sentence_keys,all_utilized_sentence_keys,support_keys,support_level,extarcted_row_for_given_id)
+    submit_button = gr.Button("Evaluate Model")
 
-print(score1)
-print(score2)
-print(score3)
+    with gr.Row():
+        with gr.Column():
+            gr.Markdown("### Output 1")
+            response = gr.Textbox(interactive=False, show_label=False, lines=2)
 
+    with gr.Row():
+        with gr.Column():
+            gr.Markdown("### Output 2")
+            output2 = gr.Textbox(interactive=False, show_label=False, lines=2)
+    
+        with gr.Column():
+            gr.Markdown("### Output 3")
+            output3 = gr.Textbox(interactive=False, show_label=False, lines=2)
 
-def chatbot(prompt):
-    return whoami()
+        with gr.Column():
+            gr.Markdown("### Output 4")
+            output4 = gr.Textbox(interactive=False, show_label=False, lines=2)
 
-iface = gr.Interface(fn=chatbot, 
-                     inputs="text", 
-                     outputs="text", 
-                     title="Capstone Project Group 10")
+    # Connect submit button to evaluation function
+    submit_button.click(EvaluateRAGModel, inputs=[question_input, dropdown_input], outputs=[response, output2, output3, output4])
 
-if __name__ == "__main__":
-    iface.launch()
+# Run the Gradio app
+iface.launch()

calculatescores.py

@@ -1,7 +1,7 @@
 import formatresultshelper
 import numpy as np
 
-from sklearn.metrics import mean_squared_error, roc_auc_score
+from sklearn.metrics import  roc_auc_score
 
 #Defined as utilized documents / retrieved documents for the query
 def compute_context_relevance(relevant_sentences, support_keys):

model.py

@@ -1,12 +0,0 @@
-from transformers import pipeline
-
-def load_model():
-    """Loads the model from Hugging Face."""
-    model = pipeline("text-generation", model="gpt2")  # Replace with your model
-    return model
-
-def generate_response(prompt):
-    """Generates a response using the model."""
-    model = load_model()
-    response = model(prompt, max_length=100, do_sample=True)
-    return response[0]["generated_text"]