squished some bugs

backend/config.json

@@ -3,8 +3,8 @@
     "llm_prompt_template" : "The following is a set of documents {docs}. Based on these docs, please summarize the content concisely. Also give a list of main concepts found in the documents. Do not add any new information. Helpful Answer: ",
     "num_return_documents" : 50,
     "embedding_model": "BAAI/bge-large-en-v1.5",
-    "llm_model": "qwen2:1.5b",
-    "num_documents_for_llm" : 10,
+    "llm_model": "llama3",
+    "num_documents_for_llm" : 50,
     "data_dir": "./data/",
     "persist_dir": "./data/chroma_db",
     "testing_flag": false,

backend/modules/llm.py

@@ -208,7 +208,7 @@ def create_vector_store(
         # subset the data for testing
         if config["test_subset_2000"] == True:
             print("[INFO] Subsetting the data to 100 rows.")
-            documents = documents[:100]
+            documents = documents[:500]
     unique_docs, unique_ids = generate_unique_documents(documents, db)
 
     print(

backend/modules/metadata_utils.py

@@ -221,7 +221,7 @@ def get_all_metadata_from_openml(config: dict) -> Tuple[pd.DataFrame, Sequence[i
         # subset the data for testing
         if config["test_subset_2000"] == True:
             print("[INFO] Subsetting the data to 100 rows.")
-            all_objects = all_objects[:100]
+            all_objects = all_objects[:500]
 
         data_id = [int(all_objects.iloc[i]["did"]) for i in range(len(all_objects))]
 
@@ -232,7 +232,7 @@ def get_all_metadata_from_openml(config: dict) -> Tuple[pd.DataFrame, Sequence[i
         openml_data_object = handler.get_metadata(data_id)
 
         print("[INFO] Saving metadata to file.")
-        save_metadata_to_file((openml_data_object, data_id, all_objects), save_filename)
+        save_metadata_to_file((openml_data_object, data_id, all_objects, handler), save_filename)
 
         return openml_data_object, data_id, all_objects, handler
 

docs/developer tutorials/train and evaluate models.ipynb

@@ -75,7 +75,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "list_of_embedding_models = [\"BAAI/bge-small-en-v1.5\"]\n",
+    "# list_of_embedding_models = [\"BAAI/bge-small-en-v1.5\", \"Alibaba-NLP/gte-Qwen2-1.5B-instruct\"]\n",
+    "list_of_embedding_models = [\"GritLM/GritLM-7B\"]\n",
     "list_of_llm_models = [\"qwen2:1.5b\", \"phi3\"]"
    ]
   },
@@ -212,6 +213,8 @@
     "        # TODO : Replace this evaluation with a more meaningful one\n",
     "        combined_df = aggregate_multiple_queries_and_count(queries,qa_dataset=qa_dataset, config=config, group_cols = [\"id\", \"name\"], sort_by=\"query\", count = True)\n",
     "\n",
+    "        # TODO : ADD LLM evaluation here when the function is ready\n",
+    "\n",
     "        combined_df.to_csv(experiment_path / \"results.csv\")"
    ]
   },
@@ -239,7 +242,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.9.19"
+   "version": "3.10.14"
   }
  },
  "nbformat": 4,

frontend/ui.py

@@ -9,7 +9,8 @@
 st.title("OpenML AI Search")
 
 query_type = st.selectbox("Select Query Type", ["Dataset", "Flow"])
-query = st.chat_input("Enter your query")
+# query = st ("Enter your query")
+query = st.text_input("Enter your query")
 
 st.session_state["query"] = query
 

ollama/get_ollama.sh

@@ -6,5 +6,5 @@ while [ "$(ollama list | grep 'NAME')" == "" ]; do
   sleep 1
 done
 
-ollama run qwen2:1.5b
+ollama run llama3
 

start_training.sh

@@ -0,0 +1,3 @@
+cd backend
+python training.py
+cd ..