-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmain_10times.py
257 lines (226 loc) · 11.6 KB
/
main_10times.py
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
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
import time
import torch
import numpy as np
import os
from scripts.model_builder import get_default_spec, save_model, load_model_only_inference
from scripts.transformer_prediction_interface import transformer_predict, get_params_from_config, TabPFNClassifier
from scripts.differentiable_pfn_evaluation import eval_model, eval_model_range
from sklearn.ensemble import BaggingClassifier,GradientBoostingClassifier
from boost_tabpfn_noweight import SamplingAdaboost_noweight
from datasets import load_openml_list, open_cc_dids, open_cc_valid_dids, test_dids_classification
import torch
from scripts import tabular_metrics
import random
import argparse
from boost_tabpfn import SamplingAdaboost
def get_args():
parser = argparse.ArgumentParser(description="None")
parser.add_argument("--gpu", default=0,type=int,help="gpu index")
parser.add_argument("--seed", default=0,type=int,help="seed")
parser.add_argument("--modelname", default="tabpfn",type=str,help="modelname")#choice=["tabpfn","bagging_tabpfn","adaboost_tabfpn"]
parser.add_argument("--ensemble_num", default=10,type=int,help="ensemble")
parser.add_argument("--sampling_size", default=0.5,type=float,help="sampling size")
parser.add_argument("--replacement",action="store_true",default=False, help="whether replacement when sampling. False =不放回")
parser.add_argument("--batch_size", default=64,type=int,help="batchsize for minibatch training")
parser.add_argument("--train_portion", default=1,type=float,help="training portion")
parser.add_argument("--loss", default="CE",type=str,help="loss for gboost")#loss for gboost## MSE,CE
parser.add_argument("--updating", default="hadamard",type=str,help="weight updating for gboost ")#updating for gboost ##adaboost,hadamard,exphadamard
parser.add_argument("--wl_num", default=2,type=int,help="number of weaklearners sampled in each round, only used for V4")
parser.add_argument("--debug",default=False, action = 'store_true')
args = parser.parse_args()
return args
def run_onetime(test_datasets,args,run=0):
ensemblenum=args.ensemble_num
datasetnames=[]
shapes=[]
rocs=[]
ces=[]
accs=[]
modelname=args.modelname
datasetportion=args.sampling_size
if args.debug:
debugresults="debug/smallresults.txt"
print(f"#################################running model {modelname} at run {run}#################################")
for i in range(len(test_datasets)):
evaluation_dataset_index = i # Index of the dataset to predict
ds = test_datasets[evaluation_dataset_index]
print(f'Evaluation dataset name: {ds[0]} shape {ds[1].shape}')
datasetnames.append(ds[0])
shapes.append(ds[1].shape)
xs, ys = ds[1].clone(), ds[2].clone()
eval_position = xs.shape[0] // 2
train_xs, train_ys = xs[0:eval_position], ys[0:eval_position]
test_xs, test_ys = xs[eval_position:], ys[eval_position:]
if modelname=="tabpfn":
classifier = TabPFNClassifier(device=device,N_ensemble_configurations=32)
classifier.fit(train_xs, train_ys)
prediction_ = classifier.predict_proba(test_xs)
elif modelname=="bagging_tabpfn":
while True:
try:
datasetportion=args.sampling_size
classifier=BaggingClassifier(TabPFNClassifier(device=device,N_ensemble_configurations=32),n_estimators=ensemblenum,max_samples=datasetportion,bootstrap=args.replacement)
classifier.fit(train_xs, train_ys)
prediction_ = classifier.predict_proba(test_xs)
break
except:
print("######## re-bagging ######")
# try:
# datasetportion=0.15
# classifier=BaggingClassifier(TabPFNClassifier(device=device,),n_estimators=ensemblenum,max_samples=datasetportion)
# classifier.fit(train_xs, train_ys)
# prediction_ = classifier.predict_proba(test_xs)
# except:
# datasetportion=0.2
# classifier=BaggingClassifier(TabPFNClassifier(device=device,),n_estimators=ensemblenum,max_samples=datasetportion)
# classifier.fit(train_xs, train_ys)
# prediction_ = classifier.predict_proba(test_xs)
elif modelname=="gboost_tabpfnV2":
from gradient_boost_tabpfn import SamplingGradientboost
datasetportion=args.sampling_size
classifier=SamplingGradientboost(TabPFNClassifier(device=device,N_ensemble_configurations=32),ensemblenum,args.sampling_size,args,version=2)
classifier.fit(train_xs, train_ys,test_xs)
prediction_ = classifier.predict_proba(test_xs)
if args.debug:
prediction_,boost_residuals,boost_loss,train_auc,test_auc,test_F_prev = classifier.predict_proba_debug(test_xs,test_ys)
debugresults="debug/smallresults"
else:
raise KeyError
roc= tabular_metrics.auc_metric(test_ys, prediction_)
try:
ce=tabular_metrics.cross_entropy(test_ys, prediction_)
except:
prediction_[prediction_>1]=1
prediction_[prediction_<0]=0
ce=tabular_metrics.cross_entropy(test_ys, prediction_)
acc=tabular_metrics.accuracy_metric(test_ys, prediction_)
# roc, ce = tabular_metrics.auc_metric(test_ys, prediction_), 0
roc=float(roc)
ce=float(ce)
# 'AUC', float(roc), 'Cross Entropy', float(ce)
print("roc:",roc," CE:",ce," ACC:",acc)
rocs.append(roc)
ces.append(ce)
accs.append(acc)
return datasetnames,rocs,ces,accs
if __name__=="__main__":
args=get_args()
base_path = '.'
max_samples = 10000
bptt = 10000
cc_test_datasets_multiclass, cc_test_datasets_multiclass_df = load_openml_list(args,open_cc_dids, multiclass=True, shuffled=True, filter_for_nan=False, max_samples = max_samples, num_feats=100, return_capped=True)
# Loading longer OpenML Datasets for generalization experiments (optional)
# test_datasets_multiclass, test_datasets_multiclass_df = load_openml_list(test_dids_classification, multiclass=True, shuffled=True, filter_for_nan=False, max_samples = 10000, num_feats=100, return_capped=True)
random.seed(0)
# def get_datasets(selector, task_type, suite='cc'):
# if task_type == 'binary':
# ds = valid_datasets_binary if selector == 'valid' else test_datasets_binary
# else:
# if suite == 'openml':
# ds = valid_datasets_multiclass if selector == 'valid' else test_datasets_multiclass
# elif suite == 'cc':
# ds = cc_valid_datasets_multiclass if selector == 'valid' else cc_test_datasets_multiclass
# else:
# raise Exception("Unknown suite")
# return ds
def get_datasets(selector, task_type, suite='cc'):
return cc_test_datasets_multiclass
model_string, longer, task_type = '', 1, 'multiclass'
eval_positions = [1000]
bptt = 2000
test_datasets, valid_datasets = get_datasets('test', task_type, suite='cc'), get_datasets('valid', task_type, suite='cc')
# [(i, test_datasets[i][0]) for i in range(len(test_datasets))]
if args.gpu>-1:
device=f"cuda:{args.gpu}"
else:
device="cpu"
########set random seeds
def set_seed(seed: int):
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
set_seed(args.seed)
# modelname="tabpfn"
rocss=[]
cess=[]
accss=[]
if args.debug:
datasetnames,rocs,ces,accs=run_onetime(test_datasets,args)
else:
for run in range(10):
datasetnames,rocs,ces,accs=run_onetime(test_datasets,args,run)
rocss.append(rocs)
cess.append(ces)
accss.append(accs)
def postprocess(results):
results=np.array(results)
results_avg=np.mean(results,axis=0)## avg for different datasets. a list
results_std=np.std(results,axis=0)
avg_results=np.mean(results,axis=-1)
totalavg=np.mean(avg_results)####avg for all datasets, a value
totalstd=np.std(avg_results)
return results_avg,results_std,totalavg,totalstd
results=[postprocess(rocss),postprocess(cess),postprocess(accss)]### roc,ce,acc
metricnames=["rocauc","ce","acc"]
# roc_avg,roc_std,roc_totalavg,roc_total_std=postprocess(rocss)
# ce_avg,ce_std,ce_totalavg,ce_total_std=postprocess(cess)
# acc_avg,acc_std,acc_totalavg,acc_total_std=postprocess(accss)
modelname=args.modelname
ensemblenum=args.ensemble_num
datasetportion=args.sampling_size
if modelname=="tabpfn":
resultname="results/tabpfn_baseline_10times.txt"
elif modelname=="bagging_tabpfn":
resultname=f"results/baggingp_tabpfn_{datasetportion}x{ensemblenum}_10times.txt"
elif modelname=="minibatch_tabpfn":
resultname=f"results/{args.modelname}_{args.batch_size}x{args.train_portion}_10times.txt"
elif modelname.startswith("gboost_tabpfn"):
resultname=f"results/{args.modelname}_{datasetportion}x{ensemblenum}_{args.loss}x{args.updating}_10times.txt"
if "4" in modelname:
resultname=f"results/{args.modelname}_{datasetportion}x{ensemblenum}_{args.loss}x{args.updating}_wl={args.wl_num}_10times.txt"
else:
resultname=f"results/{args.modelname}_{datasetportion}x{ensemblenum}_10times.txt"
with open(resultname,"w") as f:
f.write("| |")
for datasetname in datasetnames:
f.write(f" {datasetname} |")
f.write(f" Total Avg |")
f.write("\n")
f.write("| -------- |")
for datasetname in datasetnames:
f.write(" -------- |")
f.write("| -------- |")
f.write("\n")
# if modelname=="tabpfn":
# f.write(f"|{modelname}|")
# elif modelname=="minibatch_tabpfn":
# f.write(f"|{modelname},{args.batch_size},{args.train_portion}|")
# else:
# f.write(f"|{modelname},{datasetportion}x{ensemblenum}|")
for i in range(len(results)):##roc,ce,acc
result=results[i]
metricname=metricnames[i]
if modelname=="tabpfn":
f.write(f"|{modelname},{metricname}|")
elif modelname=="minibatch_tabpfn":
f.write(f"|{modelname},{args.batch_size},{args.train_portion},{metricname}|")
else:
f.write(f"|{modelname},{datasetportion}x{ensemblenum},{metricname}|")
for i in range(len(result[0])):
avg=result[0][i]
f.write(" {:.4} |".format(avg))
f.write(" {:.4} |".format(result[2]))
f.write("\n")
if modelname=="tabpfn":
f.write(f"|std-{metricname}|")
elif modelname=="minibatch_tabpfn":
f.write(f"|std-{metricname}|")
else:
f.write(f"|std-{metricname}|")
for i in range(len(result[0])):
std=result[1][i]
f.write("{:.4}|".format(std))
f.write(" {:.4}|".format(result[3]))
f.write("\n")
print(f"saved results for {modelname}")