Upload folder using huggingface_hub

app/draw_diagram.py

@@ -55,23 +55,39 @@ def draw(folder_name, category_name, dataset_name, metrics, cus_sort=True):
 
     models = st.multiselect("Please choose the model", 
                             sorted(chart_data['model_show'].tolist()), 
-                            default = 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=[new_dataset_name], ascending=cus_sort).dropna(axis=0)
 
-    # import pdb
-    # pdb.set_trace()
-
     if len(chart_data) == 0:
         return
+
+    # Get Values
+    data_values = chart_data.iloc[:, 1]
     
-    min_value = round(min(chart_data.iloc[:, 1]) - 0.1*min(chart_data.iloc[:, 1]), 1) 
-    max_value = round(max(chart_data.iloc[:, 1]) + 0.1*max(chart_data.iloc[:, 1]), 1)         
+    # Calculate Q1 and Q3
+    q1 = data_values.quantile(0.25)
+    q3 = data_values.quantile(0.75)
+
+    # Calculate IQR
+    iqr = q3 - q1
+
+    # Define lower and upper bounds (1.5*IQR is a common threshold)
+    lower_bound = q1 - 1.5 * iqr
+    upper_bound = q3 + 1.5 * iqr
+
+    # Filter data within the bounds
+    filtered_data = data_values[(data_values >= lower_bound) & (data_values <= upper_bound)]
+
+    # Calculate min and max values after outlier handling
+    min_value = round(filtered_data.min() - 0.1 * filtered_data.min(), 3)
+    max_value = round(filtered_data.max() + 0.1 * filtered_data.max(), 3)
 
     options = {
-        "title": {"text": f"{display_names[folder_name.upper()]}"},
+        #"title": {"text": f"{display_names[folder_name.upper()]}"},
+        "title": {"text": f"{dataset_name}"},
         "tooltip": {
             "trigger": "axis",
             "axisPointer": {"type": "cross", "label": {"backgroundColor": "#6a7985"}},
@@ -120,7 +136,7 @@ def draw(folder_name, category_name, dataset_name, metrics, cus_sort=True):
     # st.divider()
     with st.container():
         # st.write("")
-        st.markdown('##### TABLE')
+        st.markdown('##### Results')
         # custom_css = """
                    
         #             """
@@ -130,16 +146,46 @@ def draw(folder_name, category_name, dataset_name, metrics, cus_sort=True):
 
         chart_data['model_link'] = chart_data['model_show'].map(model_link) 
 
-        # import pdb
-        # pdb.set_trace()
-
         chart_data_table = chart_data[['model_show', chart_data.columns[1], chart_data.columns[3]]]
 
+        cur_dataset_name = chart_data_table.columns[1]
+        print(cur_dataset_name)
+
+        if cur_dataset_name in [
+                            'librispeech_test_clean',
+                            'librispeech_test_other',
+                            'common_voice_15_en_test',
+                            'peoples_speech_test',
+                            'gigaspeech_test',
+                            'earnings21_test',
+                            'earnings22_test',
+                            'tedlium3_test',
+                            'tedlium3_long_form_test',
+                            'imda_part1_asr_test',
+                            'imda_part2_asr_test',
+
+                            'aishell_asr_zh_test',
+                            ]:
+            
+            styled_df = chart_data_table.style.highlight_min(
+                subset=[chart_data_table.columns[1]], color='yellow'
+            )
+        else:
+
+            chart_data_table = chart_data_table.sort_values(
+                    by=chart_data_table.columns[1],
+                    ascending=False
+                ).reset_index(drop=True)
+
+            styled_df = chart_data_table.style.highlight_max(
+                subset=[chart_data_table.columns[1]], color='yellow'
+            )
+
         st.dataframe(
-                chart_data_table,
+                styled_df,
                 column_config={
                     'model_show': 'Model',
-                    chart_data_table.columns[1]: {'alignment': 'center'},
+                    chart_data_table.columns[1]: {'alignment': 'left'},
                     "model_link": st.column_config.LinkColumn(
                         "Model Link",
                         # # # help="",
@@ -151,6 +197,9 @@ def draw(folder_name, category_name, dataset_name, metrics, cus_sort=True):
                 hide_index=True,
                 use_container_width=True
             )
+        
+
+        
 
 
         # s = ''
@@ -210,8 +259,8 @@ def draw(folder_name, category_name, dataset_name, metrics, cus_sort=True):
     '''
     show samples
     '''
-    if dataset_name in ['Earnings21-Test', 'Earnings22-Test', 'Tedlium3-Long-form-Test']:
+    if dataset_name in ['Earnings21-Test', 'Earnings22-Test', 'Tedlium3-Test', 'Tedlium3-Long-form-Test']:
         pass
     else:
-        show_examples(category_name, dataset_name, chart_data['Model'].tolist())
+        show_examples(category_name, dataset_name, chart_data['Model'].tolist(), display_model_names)
     

app/show_examples.py

@@ -2,7 +2,7 @@ import streamlit as st
 import datasets
 import numpy as np
 
-def show_examples(category_name, dataset_name, model_lists):
+def show_examples(category_name, dataset_name, model_lists, display_model_names):
     st.divider()
     sample_folder = f"./examples/{category_name}/{dataset_name}"
     
@@ -10,7 +10,7 @@ def show_examples(category_name, dataset_name, model_lists):
 
     for index in range(len(dataset)):
         with st.container():
-            st.markdown(f'##### EXAMPLE {index+1}')
+            st.markdown(f'##### Example-{index+1}')
             col1, col2 = st.columns([0.3, 0.7], vertical_alignment="center")
 
             # with col1:
@@ -83,10 +83,11 @@ def show_examples(category_name, dataset_name, model_lists):
             with st.container():
                 custom_css = """
                             <style>
-                            .my-container-table {
-                            background-color: #cad8e7;
+                            .my-container-table, p.my-container-text {
+                            background-color: #fcf8dc;
                             padding: 10px;
                             border-radius: 5px;
+                            font-size: 13px;
                             # height: 50px;
                             word-wrap: break-word
                             }
@@ -97,16 +98,22 @@ def show_examples(category_name, dataset_name, model_lists):
                 model_lists.sort()
 
                 s = f"""<tr>
-                       <td>Reference</td>
-                       <td>{question_text}</td>
-                       <td>{dataset[index]['answer']['text']}</td>
-                </tr>"""
+                       <td><b>REFERENCE</td>
+                       <td><b>{question_text.replace('(A)', '<br>(A)').replace('(B)', '<br>(B)').replace('(C)', '<br>(C)')}
+                       </td>
+                       <td><b>{dataset[index]['answer']['text']}
+                       </td>
+                </tr>
+                """
                 if dataset_name in ['CN-College-Listen-MCQ-Test', 'DREAM-TTS-MCQ-Test']:
                     for model in model_lists:
                         try:
                             s += f"""<tr>
-                                <td>{model}</td>
-                                <td><p>{dataset[index][model]['text']}</p> <p>{choices_text}</p></td>
+                                <td>{display_model_names[model]}</td>
+                                <td>
+                                    {dataset[index][model]['text'].replace('Choices:', '<br>Choices:').replace('(A)', '<br>(A)').replace('(B)', '<br>(B)').replace('(C)', '<br>(C)') 
+                                     }
+                                    </td>
                                 <td>{dataset[index][model]['model_prediction']}</td>
                             </tr>"""
                         except:
@@ -116,15 +123,14 @@ def show_examples(category_name, dataset_name, model_lists):
                     for model in model_lists:
                         try:
                             s += f"""<tr>
-                                <td>{model}</td>
+                                <td>{display_model_names[model]}</td>
                                 <td>{dataset[index][model]['text']}</td>
                                 <td>{dataset[index][model]['model_prediction']}</td>
                             </tr>"""
                         except:
                             print(f"{model} is not in {dataset_name}")
                             continue
-                # import pdb
-                # pdb.set_trace()
+
                 
                 body_details = f"""<table style="table-layout: fixed; width:100%">
                 <thead>