import streamlit as st import pandas as pd import numpy as np from streamlit_echarts import st_echarts from streamlit.components.v1 import html import pandas as pd from model_information import get_dataframe info_df = get_dataframe() def draw(folder_name, category_one, category_two, sort, num_sort, model_size_range): folder = f"./results/{folder_name}/" data_path = f'{folder}/{category_one}/{category_two}.csv' chart_data = pd.read_csv(data_path).dropna(axis='columns').round(3) st.markdown(""" """, 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'])} model2sizes = {key.strip() :val.strip() for key, val in zip(info_df['Original Name'], info_df['Model Size'])} chart_data['model_show'] = chart_data['Model'].map(display_model_names) chart_data['model_show'] = chart_data['model_show'].fillna(chart_data['Model'].apply(lambda x: x.replace('_', '-'))) chart_data['model_size'] = chart_data['Model'].map(model2sizes) chart_data['model_size'] = chart_data['model_size'].fillna('99999') # How to work on the model size range, filter the ones that are not in the range if model_size_range != 'All': if model_size_range == '<10B': chart_data = chart_data[chart_data['model_size'].astype(float) < 10] elif model_size_range == '10B-30B': chart_data = chart_data[(chart_data['model_size'].astype(float) >= 10) & (chart_data['model_size'].astype(float) < 30)] elif model_size_range == '>30B': chart_data = chart_data[chart_data['model_size'].astype(float) >= 30] chart_data.drop(columns=['model_size'], inplace=True) models = st.multiselect("Please choose the model", sorted(chart_data['model_show'].tolist()), default = sorted(chart_data['model_show'].tolist()), ) # if 'Select All' in st.session_state.models: # st.session_state.models = chart_data['model_show'].tolist() chart_data = chart_data[chart_data['model_show'].isin(models)] if len(chart_data) == 0: return 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) display_names = { 'cross_mmlu' : 'Cross-MMLU', 'cross_mmlu_no_prompt' : 'Cross-MMLU-No-Prompt', 'cross_logiqa' : 'Cross-LogiQA', 'cross_logiqa_no_prompt' : 'Cross-LogiQA-No-Prompt', 'cross_xquad' : 'Cross-XQUAD', 'cross_xquad_no_prompt' : 'Cross-XQUAD-No-Prompt', 'sg_eval' : 'SG EVAL', 'sg_eval_v1_cleaned' : 'SG EVAL V1 Cleaned', 'sg_eval_v2_mcq' : 'SG EVAL V2 MCQ', 'sg_eval_v2_mcq_no_prompt': 'SG EVAL V2 MCQ No Prompt', 'sg_eval_v2_open' : 'SG EVAL V2 Open Ended', 'us_eval' : 'US EVAL', 'cn_eval' : 'CN EVAL', 'ph_eval' : 'PH EVAL' } data_columns = [i for i in chart_data.columns if i not in ['Model', 'model_show']] ''' Show Table ''' with st.container(): st.markdown('##### TABLE') model_link = {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 = chart_data[['model_show', 'model_link'] + data_columns] # Format numeric columns to 2 decimal places chart_data_table[chart_data_table.columns[2]] = chart_data_table[chart_data_table.columns[2]].apply(lambda x: round(float(x), 3) if isinstance(x, (int, float)) else x) chart_data_table = chart_data_table.sort_values( by=chart_data_table.columns[2], ascending=False ).reset_index(drop=True) styled_df = chart_data_table.style.format( { chart_data_table.columns[i]: "{:.3f}" for i in range(2, len(chart_data_table.columns)) } ).highlight_max( subset=[chart_data_table.columns[2]], color='yellow' ) st.dataframe( styled_df, column_config={ 'model_show': 'Model', chart_data_table.columns[1]: {'alignment': 'center'}, "model_link": st.column_config.LinkColumn( "Model Link", ), }, hide_index=True, use_container_width=True ) # = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = # Initialize a session state variable for toggling the chart visibility if "show_chart" not in st.session_state: st.session_state.show_chart = False # Create a button to toggle visibility if st.button("Show Chart"): st.session_state.show_chart = not st.session_state.show_chart if st.session_state.show_chart: with st.container(): st.markdown('##### CHART') if num_sort == 'Ascending': ascend = True else: ascend = False chart_data = chart_data.sort_values(by=[sort], ascending=ascend).dropna(axis=0) options = { # "title": {"text": f"{display_names[category_two]}"}, "tooltip": { "trigger": "axis", "axisPointer": {"type": "cross", "label": {"backgroundColor": "#6a7985"}}, "triggerOn": 'mousemove', }, "legend": {"data": data_columns}, "toolbox": {"feature": {"saveAsImage": {}}}, "grid": {"left": "3%", "right": "4%", "bottom": "3%", "containLabel": True}, "xAxis": [ { "type": "category", "boundaryGap": True, "triggerEvent": True, "data": chart_data['model_show'].tolist(), } ], "yAxis": [{"type": "value", "min": min_value, "max": max_value, "boundaryGap": True # "splitNumber": 10 }], "series": [{ "name": f"{col}", "type": "bar", "data": chart_data[f'{col}'].tolist(), } for col in data_columns], } events = { "click": "function(params) { return params.value }" } value = st_echarts(options=options, events=events, height="500px")