Add initial data files and Streamlit app for CNO classification

- Created defaults_session_state.json to manage session state defaults.
- Added idxs.csv containing CNO labels and indices.
- Introduced metadata.json with demographic and educational metadata.
- Added problematic_cnos.json listing unreliable CNO codes.
- Included model.safetensors and sentencepiece.bpe.model for model loading.
- Created special_tokens_map.json for token management in the model.
- Developed streamlit_app.py for user interface and interaction.
- Added tokenizer.json and tokenizer_config.json for tokenizer settings.
- Included training_args.bin for model training arguments.
- Implemented utility functions in cno_utils.py for label and description mapping.

src/streamlit_app.py → streamlit_app.py

@@ -73,8 +73,8 @@ def load_json_file(filename):
 
 
 pipe = load_huggingface_model()
-METADATA = load_json_file("src/data/metadata.json")
-PROBLEMATIC_CNOS = load_json_file("src/data/problematic_cnos.json")
+METADATA = load_json_file("data/metadata.json")
+PROBLEMATIC_CNOS = load_json_file("data/problematic_cnos.json")
 
 
 def run_inference(text_input):
@@ -117,7 +117,7 @@ st.sidebar.markdown(
 
 
 def init_state() -> None:
-    defaults = load_json_file("src/data/defaults_session_state.json")
+    defaults = load_json_file("data/defaults_session_state.json")
     for k, v in defaults.items():
         st.session_state.setdefault(k, v)