Spaces:
Runtime error
Runtime error
File size: 4,957 Bytes
a6e43e6 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 |
import copy as cp
import json
from collections import defaultdict
from urllib.request import urlopen
import gradio as gr
import numpy as np
import pandas as pd
from meta_data import META_FIELDS, URL
def listinstr(lst, s):
assert isinstance(lst, list)
for item in lst:
if item in s:
return True
return False
def load_results():
data = json.loads(urlopen(URL).read())
return data
def nth_large(val, vals):
return sum([1 for v in vals if v > val]) + 1
def format_timestamp(timestamp):
date = timestamp[:2] + '.' + timestamp[2:4] + '.' + timestamp[4:6]
time = timestamp[6:8] + ':' + timestamp[8:10] + ':' + timestamp[10:12]
return date + ' ' + time
def model_size_flag(sz, FIELDS):
if pd.isna(sz) and 'Unknown' in FIELDS:
return True
if pd.isna(sz):
return False
if '<10B' in FIELDS and sz < 10:
return True
if '10B-20B' in FIELDS and sz >= 10 and sz < 20:
return True
if '20B-40B' in FIELDS and sz >= 20 and sz < 40:
return True
if '>40B' in FIELDS and sz >= 40:
return True
return False
def model_type_flag(line, FIELDS):
if 'OpenSource' in FIELDS and line['OpenSource'] == 'Yes':
return True
if 'API' in FIELDS and line['OpenSource'] == 'No' and line['Verified'] == 'Yes':
return True
if 'Proprietary' in FIELDS and line['OpenSource'] == 'No' and line['Verified'] == 'No':
return True
return False
def BUILD_L1_DF(results, fields):
res = defaultdict(list)
for i, m in enumerate(results):
item = results[m]
meta = item['META']
for k in META_FIELDS:
if k == 'Parameters (B)':
param = meta['Parameters']
res[k].append(float(param.replace('B', '')) if param != '' else None)
elif k == 'Method':
name, url = meta['Method']
res[k].append(f'<a href="{url}">{name}</a>')
else:
res[k].append(meta[k])
scores, ranks = [], []
for d in fields:
res[d].append(item[d]['Overall'])
if d == 'MME':
scores.append(item[d]['Overall'] / 28)
else:
scores.append(item[d]['Overall'])
ranks.append(nth_large(item[d]['Overall'], [x[d]['Overall'] for x in results.values()]))
res['Avg Score'].append(round(np.mean(scores), 1))
res['Avg Rank'].append(round(np.mean(ranks), 2))
df = pd.DataFrame(res)
df = df.sort_values('Avg Rank')
check_box = {}
check_box['essential'] = ['Method', 'Parameters (B)', 'Language Model', 'Vision Model']
check_box['required'] = ['Avg Score', 'Avg Rank']
check_box['all'] = check_box['required'] + ['OpenSource', 'Verified'] + fields
type_map = defaultdict(lambda: 'number')
type_map['Method'] = 'html'
type_map['Language Model'] = type_map['Vision Model'] = type_map['OpenSource'] = type_map['Verified'] = 'str'
check_box['type_map'] = type_map
return df, check_box
def BUILD_L2_DF(results, dataset):
res = defaultdict(list)
fields = list(list(results.values())[0][dataset].keys())
non_overall_fields = [x for x in fields if 'Overall' not in x]
overall_fields = [x for x in fields if 'Overall' in x]
if dataset == 'MME':
non_overall_fields = [x for x in non_overall_fields if not listinstr(['Perception', 'Cognition'], x)]
overall_fields = overall_fields + ['Perception', 'Cognition']
for m in results:
item = results[m]
meta = item['META']
for k in META_FIELDS:
if k == 'Parameters (B)':
param = meta['Parameters']
res[k].append(float(param.replace('B', '')) if param != '' else None)
elif k == 'Method':
name, url = meta['Method']
res[k].append(f'<a href="{url}">{name}</a>')
else:
res[k].append(meta[k])
fields = [x for x in fields]
for d in non_overall_fields:
res[d].append(item[dataset][d])
for d in overall_fields:
res[d].append(item[dataset][d])
df = pd.DataFrame(res)
all_fields = overall_fields + non_overall_fields
# Use the first 5 non-overall fields as required fields
required_fields = overall_fields if len(overall_fields) else non_overall_fields[:5]
if 'Overall' in overall_fields:
df = df.sort_values('Overall')
df = df.iloc[::-1]
check_box = {}
check_box['essential'] = ['Method', 'Parameters (B)', 'Language Model', 'Vision Model']
check_box['required'] = required_fields
check_box['all'] = all_fields
type_map = defaultdict(lambda: 'number')
type_map['Method'] = 'html'
type_map['Language Model'] = type_map['Vision Model'] = type_map['OpenSource'] = type_map['Verified'] = 'str'
check_box['type_map'] = type_map
return df, check_box
|