Upload folder using huggingface_hub

app/content.py

@@ -145,9 +145,7 @@ dataset_diaplay_information = {
     'YTB-SQA-Batch1': 'Under Development',
     'YTB-SDS-Batch1': 'Under Development',
     'YTB-PQA-Batch1': 'Under Development',
-    
-                }
-
+    }
 
 
 

app/draw_diagram.py

@@ -1,6 +1,8 @@
 import streamlit as st
 import pandas as pd
 import numpy as np
+import json
+
 from streamlit_echarts import st_echarts
 from app.show_examples import *
 from app.content import *
@@ -11,47 +13,56 @@ from model_information import get_dataframe
 info_df = get_dataframe()
 
 
-def draw(folder_name, category_name, displayname, metrics, cus_sort=True):
+def draw_table(dataset_displayname, metrics):
+
+    dataset_nickname = displayname2datasetname[dataset_displayname]
+
+    with open('organize_model_results.json', 'r') as f:
+        organize_model_results = json.load(f)
+
+    model_results      = organize_model_results[dataset_nickname][metrics]
+    model_name_mapping = {key.strip(): val for key, val in zip(info_df['Original Name'], info_df['Proper Display Name'])}
+    model_results      = {model_name_mapping.get(key, key): val for key, val in model_results.items()}
+
     
-    folder = f"./results_organized/{metrics}/"
+    # folder = f"./results_organized/{metrics}/"
 
-    # Load the results from CSV
-    data_path = f'{folder}/{category_name.lower()}.csv'
-    chart_data = pd.read_csv(data_path).round(3)
+    # # Load the results from CSV
+    # data_path = f'{folder}/{category_name.lower()}.csv'
+    # chart_data = pd.read_csv(data_path).round(3)
     
-    dataset_name = displayname2datasetname[displayname]
-    chart_data = chart_data[['Model', dataset_name]]
-
-    # Rename to proper display name
-    chart_data = chart_data.rename(columns=datasetname2diaplayname)
-
-
-    st.markdown("""
-                <style>
-                .stMultiSelect [data-baseweb=select] span {
-                    max-width: 800px;
-                    font-size: 0.9rem;
-                    background-color: #3C6478 !important; /* Background color for selected items */
-                    color: white; /* Change text color */
-                    back
-                }
-                </style>
-                """, unsafe_allow_html=True)
+    # dataset_name = displayname2datasetname[displayname]
+    # chart_data = chart_data[['Model', dataset_name]]
+
+    # # Rename to proper display name
+    # chart_data = chart_data.rename(columns=datasetname2diaplayname)
+
+    # st.markdown("""
+    #             <style>
+    #             .stMultiSelect [data-baseweb=select] span {
+    #                 max-width: 800px;
+    #                 font-size: 0.9rem;
+    #                 background-color: #3C6478 !important; /* Background color for selected items */
+    #                 color: white; /* Change text color */
+    #                 back
+    #             }
+    #             </style>
+    #             """, unsafe_allow_html=True)
     
-    # remap model names
-    display_model_names = {key.strip() :val.strip() for key, val in zip(info_df['Original Name'], info_df['Proper Display Name'])}
-    chart_data['model_show'] = chart_data['Model'].map(lambda x: display_model_names.get(x, x))
+    # # remap model names
+    # display_model_names = {key.strip() :val.strip() for key, val in zip(info_df['Original Name'], info_df['Proper Display Name'])}
+    # chart_data['model_show'] = chart_data['Model'].map(lambda x: display_model_names.get(x, x))
 
 
-    models = st.multiselect("Please choose the model", 
-                            sorted(chart_data['model_show'].tolist()), 
-                            default = sorted(chart_data['model_show'].tolist()),
-                            )
+    # models = st.multiselect("Please choose the model", 
+    #                         sorted(chart_data['model_show'].tolist()), 
+    #                         default = sorted(chart_data['model_show'].tolist()),
+    #                         )
     
-    chart_data = chart_data[chart_data['model_show'].isin(models)]
-    chart_data = chart_data.sort_values(by=[displayname], ascending=cus_sort).dropna(axis=0)
+    # chart_data = chart_data[chart_data['model_show'].isin(models)]
+    # chart_data = chart_data.sort_values(by=[displayname], ascending=cus_sort).dropna(axis=0)
 
-    if len(chart_data) == 0: return
+    # if len(chart_data) == 0: return
 
 
 
@@ -62,28 +73,27 @@ def draw(folder_name, category_name, displayname, metrics, cus_sort=True):
     with st.container():
         st.markdown('##### TABLE')
 
-        
-        model_link = {key.strip(): val for key, val in zip(info_df['Proper Display Name'], info_df['Link'])}
+        model_link_mapping = {key.strip(): val for key, val in zip(info_df['Proper Display Name'], info_df['Link'])}
 
-        chart_data['model_link'] = chart_data['model_show'].map(model_link) 
+        chart_data_table = pd.DataFrame(list(model_results.items()), columns=["model_show", dataset_displayname])
+        chart_data_table["model_link"] = chart_data_table["model_show"].map(model_link_mapping)
 
-        chart_data_table = chart_data[['model_show', chart_data.columns[1], chart_data.columns[3]]]
+        # chart_data['model_link'] = chart_data['model_show'].map(model_link) 
+
+        # chart_data_table = chart_data[['model_show', chart_data.columns[1], chart_data.columns[3]]]
 
         # Format numeric columns to 2 decimal places
         #chart_data_table[chart_data_table.columns[1]] = chart_data_table[chart_data_table.columns[1]].apply(lambda x: round(float(x), 3) if isinstance(float(x), (int, float)) else float(x))
-        cur_dataset_name = chart_data_table.columns[1]
+        # dataset_name = chart_data_table.columns[1]
 
 
         def highlight_first_element(x):
                 # Create a DataFrame with the same shape as the input
                 df_style = pd.DataFrame('', index=x.index, columns=x.columns)
-                # Apply background color to the first element in row 0 (df[0][0])
-                # df_style.iloc[0, 1] = 'background-color: #b0c1d7; color: white'
                 df_style.iloc[0, 1] = 'background-color: #b0c1d7'
-                
                 return df_style
 
-        if cur_dataset_name in [
+        if dataset_displayname in [
                             'LibriSpeech-Clean',
                             'LibriSpeech-Other',
                             'CommonVoice-15-EN',
@@ -136,11 +146,9 @@ def draw(folder_name, category_name, displayname, metrics, cus_sort=True):
         st.dataframe(
                 styled_df,
                 column_config={
-                    'model_show': 'Model',
+                    'model_show'               : 'Model',
                     chart_data_table.columns[1]: {'alignment': 'left'},
-                    "model_link": st.column_config.LinkColumn(
-                        "Model Link",
-                    ),
+                    "model_link"               : st.column_config.LinkColumn("Model Link"),
                 },
                 hide_index=True,
                 use_container_width=True
@@ -166,7 +174,7 @@ def draw(folder_name, category_name, displayname, metrics, cus_sort=True):
             st.markdown('##### CHART')
 
             # Get Values
-            data_values = chart_data.iloc[:, 1]
+            data_values = chart_data_table.iloc[:, 1]
             
             # Calculate Q1 and Q3
             q1 = data_values.quantile(0.25)
@@ -201,7 +209,7 @@ def draw(folder_name, category_name, displayname, metrics, cus_sort=True):
                         "type": "category",
                         "boundaryGap": True,
                         "triggerEvent": True,
-                        "data":  chart_data['model_show'].tolist(),
+                        "data":  chart_data_table['model_show'].tolist(),
                     }
                 ],
                 "yAxis": [{"type": "value", 
@@ -211,9 +219,9 @@ def draw(folder_name, category_name, displayname, metrics, cus_sort=True):
                             # "splitNumber": 10
                             }],
                 "series": [{
-                        "name": f"{dataset_name}",
+                        "name": f"{dataset_nickname}",
                         "type": "bar",
-                        "data": chart_data[f'{displayname}'].tolist(),
+                        "data": chart_data_table[f'{dataset_displayname}'].tolist(),
                     }],
             }
             
@@ -242,7 +250,6 @@ def draw(folder_name, category_name, displayname, metrics, cus_sort=True):
         st.session_state.show_examples = not st.session_state.show_examples
 
     if st.session_state.show_examples:
-        
         st.markdown('To be implemented')
 
         # # if dataset_name in ['Earnings21-Test', 'Earnings22-Test', 'Tedlium3-Test', 'Tedlium3-Long-form-Test']:

app/pages.py

@@ -4,7 +4,6 @@ from app.content import *
 from app.summarization import *
 
 def dataset_contents(dataset, metrics):
-    
     custom_css = """
                 <style>
                 .my-dataset-info {
@@ -39,7 +38,6 @@ def dashboard():
             **Resource for AudioLLMs:** [![GitHub Repo stars](https://img.shields.io/github/stars/AudioLLMs/Awesome-Audio-LLM?style=social)][gh2]
             """)
 
-
     st.markdown("""
             #### Recent updates
             - **Jan. 2025**: AudioBench is officially accepted to NAACL 2025!
@@ -51,7 +49,6 @@ def dashboard():
             """)
 
     st.divider()
-    
     st.markdown("""
                 #### Evaluating Audio-based Large Language Models
                 
@@ -62,9 +59,7 @@ def dashboard():
                 """
                 )
 
-
-    with st.container():
-        
+    with st.container():        
         st.markdown('''
                 ''')
         
@@ -113,15 +108,9 @@ def dashboard():
                             year={2024}
                             }
                         ```
-
                         """)
 
 
-
-
-
-
-
 def asr_english():
     st.title("Task: Automatic Speech Recognition - English")
     
@@ -143,15 +132,14 @@ def asr_english():
     left, center, _, middle, right = st.columns([0.4, 0.2, 0.2, 0.2 ,0.2])
     
     with left:
-        filter_1 = st.selectbox('Dataset', filters_levelone)
+        tab_section = st.selectbox('Dataset', filters_levelone)
     
-    if filter_1:
-        if filter_1 in sum:
+    if tab_section:
+        if tab_section in sum:
             sum_table_mulit_metrix('asr_english', ['wer'])
         else:
-            dataset_contents(dataset_diaplay_information[filter_1], metrics_info['wer'])
-            draw('su', 'asr_english', filter_1, 'wer', cus_sort=True)
-
+            dataset_contents(dataset_diaplay_information[tab_section], metrics_info['wer'])
+            draw_table(tab_section, 'wer')