/
test.py
161 lines (119 loc) · 6.25 KB
/
test.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
import os
import pandas as pd
import pickle
import argparse
import torch
import math
import string
import logging
import numpy as np
import json
from tqdm import tqdm
from torch.utils.data import TensorDataset, DataLoader, SequentialSampler
from transformers import AutoTokenizer,XGLMTokenizer
from transformers import AutoModelForCausalLM,XGLMForCausalLM
from sklearn.metrics import f1_score
import random
from ICL.data import ICLData
from ICL.model import ICLModel
from utils.data import load_data,task2lang,get_logic,uniform_label
def main(args):
print("Loding Model")
model=ICLModel()
model.load(model_name=args.model_name)
model.cuda()
model.eval()
if args.use_demonstrations:
args.max_length= min(args.max_length_per_example * args.k, args.max_length)
if args.set_up=='no_demo':
raise ValueError('Use_demonstrations in no_demo setup ??')
else:
args.max_length=args.max_length_per_example # i.e 512
args.set_up='no_demo'
seeds=args.seeds.split(',')
if args.tar_l==None:
tar_l=task2lang[args.dataset_name]
else:
tar_l=args.tar_l.split(',')
for s in seeds:
for tar in tar_l:
if args.dataset_name in ['amaz_bi','hateval','cls']:
args.options=None
else:
raise ValueError('Expected options')
src_l=task2lang[args.dataset_name].copy()
print(tar)
src_l.remove(tar)
for src in src_l:
if args.use_logic==True:
logic=get_logic(tar,src,args.dataset_name,flip=args.flip,simple=args.simple_logic)
elif args.cor_logic==True:
cor_lang=task2lang[args.dataset_name].copy()
cor_lang.remove(src)
cor_lang.remove(tar)
tar_cor=random.sample(cor_lang,k=1)[0]
print(src,tar,tar_cor)
logic=get_logic(tar_cor,src,args.dataset_name,flip=args.flip,simple=args.simple_logic)
else:
logic=None
print("Evaluating tar {} src {} seed {}".format(tar,src,s))
test_data=load_data(dataset_name=args.dataset_name,tar=tar,src=src,set_up=args.set_up,seed=s,k=args.k)
if args.label_uniform==True:
test_data=uniform_label(tar,src,test_data)
#test_data=test_data[:100]
data=ICLData(model_name=args.model_name,use_demonstrations=args.use_demonstrations,use_logic=(args.use_logic or args.cor_logic) ,k=args.k,
max_length=args.max_length,max_length_per_example=args.max_length_per_example,method=args.method)
data.tensorize(data=test_data,options=args.options,logic=logic,add_newlines=args.add_newlines)
print(data.print_tensorized_example())
#continue
predictions=model.do_predict(data=data,batch_size=args.batch_size)
groundtruths=[dp['output'] for dp in test_data]
eval_dic=data.evaluate(predictions,groundtruths)
if args.dataset_name in ['amaz_bi','hateval','cls']:
eval_dic['f1']=f1_score(groundtruths,predictions,average='macro')
eval_dic['seed']=s
eval_dic['dataset_name']=args.dataset_name
eval_dic['target_language']=tar
eval_dic['source_language']=src
eval_dic['set_up']=args.set_up
eval_dic['k']=args.k
eval_dic['logic']=(args.use_logic or args.cor_logic)
eval_dic['model_name']=args.model_name
set_up_name=args.set_up + ('_logic' if args.use_logic else '')
result_path='result_{}_{}_k={}.csv'.format(args.dataset_name,set_up_name,args.k)
if os.path.exists(result_path):
result_df=pd.read_csv(result_path)
result_df=result_df.append(eval_dic,ignore_index=True)
else:
result_df=pd.DataFrame([eval_dic])
result_df.to_csv(result_path,index=False)
if args.model_name=='facebook/xglm-7.5B':
out_path=f'output/{args.dataset_name}/{args.set_up}/{tar}'
out_file=os.path.join(out_path,f'model=7.5Bk={args.k}s={s}.json')
os.makedirs(out_path,exist_ok=True)
with open(out_file,'w') as fp:
json.dump(predictions,fp)
print("Infrence Done for seed{} tar {} src{}".format(s,tar,src))
if __name__=='__main__':
parser=argparse.ArgumentParser()
parser.add_argument('--use_demonstrations',default=False,action='store_true')
parser.add_argument('--use_logic',default=False,action='store_true')
parser.add_argument('--label_uniform',default=False,action='store_true')
parser.add_argument('--simple_logic',default=False,action='store_true')
parser.add_argument('--flip',default=False,action='store_true')
parser.add_argument('--cor_logic',default=False,action='store_true')
parser.add_argument('--add_newlines',default=False,action='store_true')
parser.add_argument('--dataset_name',type=str,required=True,help='The dataset_name')
parser.add_argument('--k',type=int,default=16)
parser.add_argument('--seeds',type=str,default='32,5,232,100,42')
parser.add_argument('--set_up', type=str, required=True,choices=['src_is_cross','sim_in_cross'])
parser.add_argument('--batch_size',type=int,default=8)
parser.add_argument('--model_name',type=str,default="facebook/xglm-564M")
parser.add_argument('--options',type=str,default=None)
parser.add_argument('--method',type=str,default='direct')
parser.add_argument('--max_length_per_example',type=int,default=512)
parser.add_argument('--max_length',type=int,default=512)
parser.add_argument('--tar_l',type=str,default=None,help='pass string of family you want to sample')
parser.add_argument('--src_l',type=str,default=None,help='pass string of family you want to sample')
args=parser.parse_args()
main(args)