first_commit

app.py

@@ -1,7 +1,9 @@
 import streamlit as st
 import pandas as pd
 import json
-
+from note import SUBMISSION
+from st_aggrid import JsCode
+from st_aggrid import AgGrid, GridOptionsBuilder
 def load_data():
     with open("data.jsonl", 'r', encoding='utf-8') as file:
         data = [json.loads(line) for line in file]
@@ -13,45 +15,123 @@ def case_insensitive_search(data, query, column):
         return data[data[column].str.lower().str.contains(query.lower())]
     return data
 
-def display_table(data, rows_per_page=10):
+def colored_note(text, background_color='#fcfced', text_color='black'):
+    st.markdown(f"""
+        <div style='background-color: {background_color}; color: {text_color}; 
+                    border-radius: 8px; padding: 10px; margin: 8px 0; box-shadow: 2px 2px 5px grey;'>
+            {text}
+            """, unsafe_allow_html=True)
+
+html_render = JsCode(
+        """
+    class UrlCellRenderer {
+      init(params) {
+        this.eGui = document.createElement('span'); 
+        if (params.value && params.value.includes('href=\\"')) {
+          const parser = new DOMParser();
+          const parsedHtml = parser.parseFromString(params.value, 'text/html');
+          const link = parsedHtml.querySelector('a');
+          if (link) {
+            this.eGui = document.createElement('a');
+            this.eGui.setAttribute('href', link.getAttribute('href'));
+            this.eGui.innerText = link.innerText; 
+            this.eGui.setAttribute('style', "text-decoration:none");
+            this.eGui.setAttribute('target', "_blank");
+          } else {
+            this.eGui.innerText = params.value; 
+          }
+        } else {
+          this.eGui.innerText = params.value;
+        }
+      }
+      getGui() {
+        return this.eGui;
+      }
+    }
+    """
+) 
+def display_table(data, rows_per_page=12):
+    st.markdown("""
+        <style>
+        .centered {
+            display: flex;
+            justify-content: center;
+        }
+        .css-1l02zno {
+            flex: 1;
+        }
+        </style>
+        """, unsafe_allow_html=True)
     container = st.container()
+    gb = GridOptionsBuilder.from_dataframe(data)
+    gb.configure_columns(['Chat Model','Embedding Model','Reranker Model','Framework'], 
+                        cellRenderer=html_render,
+                        sortable=True, filterable=True, resizable=True, )
+    gb.configure_column("Accuracy", sort='desc') 
+    gridOptions = gb.build()
     with container:
-        height = min(40 + rows_per_page * 38, 800)  
-        st.dataframe(data, height=height)
+        height = min(40 + rows_per_page * 38, 800)
+        col2, col3 = st.columns([5, 3])
+        with col2:
+            st.markdown("""
+            <style>
+                .ag-theme-balham {
+                    height: 500px; 
+                    width: 50%;
+                    margin: auto; 
+                }
+            </style>
+            """, unsafe_allow_html=True)
+            AgGrid(data, height=height, gridOptions=gridOptions, allow_unsafe_jscode=True)
+        with col3:
+            colored_note(SUBMISSION)
 
 def main():
-    st.title("Multihop-RAG Benchmark 💡")
+    st.set_page_config(layout="wide")
+    st.title("Multihop-RAG 💡")
+    st.write("Displaying results across different frameworks, embedding models, chat models, and chunks.")
 
     data = load_data()
 
-    st.sidebar.header("Search Options")
-    chat_model_query = st.sidebar.text_input("Chat Model")
-    embedding_model_query = st.sidebar.text_input("Embedding Model")
-    chunk_query = st.sidebar.text_input("Chunk") 
-    frame_query = st.sidebar.text_input("Framework") 
-
-    if chat_model_query:
-        data = case_insensitive_search(data, chat_model_query, 'chat_model')
-    if embedding_model_query:
-        data = case_insensitive_search(data, embedding_model_query, 'embedding_model')
-    if chunk_query:  
-        data = case_insensitive_search(data, chunk_query, 'chunk')
-    if frame_query:
-        data = case_insensitive_search(data, frame_query, 'framework')
+    st.markdown("""
+    <style>
+    div.stButton > button:first-child {
+        height: 2em;     
+        width: 100%;    
+        margin-top: 1.8em;
+    }
+    </style>
+    """, unsafe_allow_html=True)
+    col1, col2, col3, col4, col5 = st.columns(5)
 
-    st.write("Displaying results across different frameworks, embedding models, chat models, and chunks.")
+    with col1:
+        chat_model_query = st.text_input("Chat Model", key="chat_model_query")
+    with col2:
+        embedding_model_query = st.text_input("Embedding Model", key="embedding_model_query")
+    with col3:
+        chunk_query = st.text_input("Chunk", key="chunk_query")
+    with col4:
+        frame_query = st.text_input("Framework", key="frame_query")
+    with col5:
+        search_button = st.button("🔍 Search")
+
+    if search_button:
+        if chat_model_query:
+            data = case_insensitive_search(data, chat_model_query, 'Chat Model')
+        if embedding_model_query:
+            data = case_insensitive_search(data, embedding_model_query, 'Embedding Model')
+        if chunk_query:
+            data = case_insensitive_search(data, chunk_query, 'Chunk Size')
+        if frame_query:
+            data = case_insensitive_search(data, frame_query, 'Framework')
+            
     st.info("Retrieval Stage: MRR@10 and Hit@10; Response Stage: Accuracy ")
-    display_table(data) 
 
-    st.sidebar.header("Citation")
-    st.sidebar.info(
-        "Please cite this dataset as:\n"
-        "Tang, Yixuan, and Yi Yang. MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries. ArXiv, 2024,  /abs/2401.15391."
-    )
+    display_table(data)
     st.markdown("---")
     st.caption("For citation, please use: 'Tang, Yixuan, and Yi Yang. MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries. ArXiv, 2024,  /abs/2401.15391. '")
-    st.markdown("---")
-    st.caption("For results self-reporting, please send an email to ytangch@connect.ust.hk")
+    # st.markdown("---")
+    # st.caption("For results self-reporting, please send an email to ytangch@connect.ust.hk")
  
 if __name__ == "__main__":
     main()

data.jsonl

@@ -1,3 +1,6 @@
-{"Framework":"naive_RAG","Chat Model":"GPT-4","Embedding Model":"voyage-02","Reranker Model":"BAAI\/bge-reranker-large","Chunk Size":256,"MRR@10":0.3934,"Hit@10":0.6506,"Accuracy":0.56}
-{"Framework":"naive_RAG","Chat Model":"ChatGPT","Embedding Model":"voyage-02","Reranker Model":"BAAI\/bge-reranker-large","Chunk Size":256,"MRR@10":0.3934,"Hit@10":0.6506,"Accuracy":0.44}
-{"Framework":"naive_RAG","Chat Model":"Llama-2-70b-chat-hf","Embedding Model":"voyage-02","Reranker Model":"BAAI\/bge-reranker-large","Chunk Size":256,"MRR@10":0.3934,"Hit@10":0.6506,"Accuracy":0.28}
+{"Framework":"naive_RAG","Chat Model":"<a href=\"https://openai.com/index/gpt-4/\" target=\"_blank\">GPT-4</a>","Embedding Model":"<a href=\"https://docs.voyageai.com/docs/embeddings\" target=\"_blank\">voyage-02</a>","Reranker Model":"<a href=\"https://huggingface.co/BAAI/bge-reranker-large\" target=\"_blank\">BAAI/bge-reranker-large</a>","Chunk Size":256,"MRR@10":0.3934,"Hit@10":0.6506,"Accuracy":0.56}
+{"Framework":"naive_RAG","Chat Model":"<a href=\"https://openai.com/chatgpt/\" target=\"_blank\">ChatGPT</a>","Embedding Model":"<a href=\"https://docs.voyageai.com/docs/embeddings\" target=\"_blank\">voyage-02</a>","Reranker Model":"<a href=\"https://huggingface.co/BAAI/bge-reranker-large\" target=\"_blank\">BAAI/bge-reranker-large</a>","Chunk Size":256,"MRR@10":0.3934,"Hit@10":0.6506,"Accuracy":0.44}
+{"Framework":"naive_RAG","Chat Model":"<a href=\"https://huggingface.co/meta-llama/Llama-2-70b-chat-hf\" target=\"_blank\">meta-llama/Llama-2-70b-chat-hf</a>","Embedding Model":"<a href=\"https://docs.voyageai.com/docs/embeddings\" target=\"_blank\">voyage-02</a>","Reranker Model":"<a href=\"https://huggingface.co/BAAI/bge-reranker-large\" target=\"_blank\">BAAI/bge-reranker-large</a>","Chunk Size":256,"MRR@10":0.3934,"Hit@10":0.6506,"Accuracy":0.28}
+{"Framework":"naive_RAG","Chat Model":"<a href=\"https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1\" target=\"_blank\">mistralai\/Mixtral-8x7B-Instruct-v0.1</a>","Embedding Model":"<a href=\"https://docs.voyageai.com/docs/embeddings\" target=\"_blank\">voyage-02</a>","Reranker Model":"<a href=\"https://huggingface.co/BAAI/bge-reranker-large\" target=\"_blank\">BAAI/bge-reranker-large</a>","Chunk Size":256,"MRR@10":0.3934,"Hit@10":0.6506,"Accuracy":0.32}
+{"Framework":"naive_RAG","Chat Model":"<a href=\"https://www.anthropic.com/news/claude-2-1\" target=\"_blank\">Claude-2.1</a>","Embedding Model":"<a href=\"https://docs.voyageai.com/docs/embeddings\" target=\"_blank\">voyage-02</a>","Reranker Model":"<a href=\"https://huggingface.co/BAAI/bge-reranker-large\" target=\"_blank\">BAAI/bge-reranker-large</a>","Chunk Size":256,"MRR@10":0.3934,"Hit@10":0.6506,"Accuracy": 0.52}
+{"Framework":"naive_RAG","Chat Model":"<a href=\"https://ai.google/discover/palm2/\" target=\"_blank\">Google-PaLM</a>","Embedding Model":"<a href=\"https://docs.voyageai.com/docs/embeddings\" target=\"_blank\">voyage-02</a>","Reranker Model":"<a href=\"https://huggingface.co/BAAI/bge-reranker-large\" target=\"_blank\">BAAI/bge-reranker-large</a>","Chunk Size":256,"MRR@10":0.3934,"Hit@10":0.6506,"Accuracy": 0.47}

note.py

@@ -0,0 +1,19 @@
+
+SUBMISSION = """For the results report, please send the results to <strong>ytangch@connect.ust</strong><br><br>
+The experimental results should be a .zip file containing the test metrics and experimental records (such as the retrieved chunks and LLM QA process records).<br>
+
+<strong>Required Record:</strong>
+* readme.md
+```
+1. Framework 
+1. Chat Model
+2. Base Model
+3. Chunk Size (if no chunk, please fill with none)
+4. MRR@10,Hit@10,Accuracy
+```
+* addition support
+```
+5. Retrieved Record For Each Query (in .json)
+6. QA Record For Each Query (in .json)
+```
+"""

requirements.txt

@@ -1,2 +1,3 @@
 fuzzywuzzy
 st-gsheets-connection
+streamlit-aggrid