-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathrunning_different_configs.py
executable file
·156 lines (119 loc) · 8.79 KB
/
running_different_configs.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
from driver import ex
import random
import os
import argparse, sys
import pickle
parser=argparse.ArgumentParser()
parser.add_argument('--dataset', help='the dataset you want to work on')
dataset_specific_config = {
#Train:10569,dev:2642,Test:3303
"TED_humor":{'input_dims':[1,81,75,300],'max_seq_len':20,'dev_batch_size':2645,'test_batch_size':3305},
"TED_humor_albert":{'input_dims':[1,81,75,768],'max_seq_len':20,'dev_batch_size':2645,'test_batch_size':3305},
"mosi":{'input_dims':[300,5,20],'text_indices':(0,300),'audio_indices':(300,305),'video_indices':(305,325),'max_seq_len':20,'dev_batch_size':250,'test_batch_size':700},
"iemocap":{'text_indices':(0,300),'audio_indices':(300,374),'video_indices':(374,409),'max_seq_len':21},
"mmmo":{'text_indices':(0,300),'audio_indices':(300,374),'video_indices':(374,409),'max_seq_len':21},
"moud":{'text_indices':(0,300),'audio_indices':(300,374),'video_indices':(374,409),'max_seq_len':21},
"pom":{'text_indices':(0,300),'audio_indices':(300,343),'video_indices':(343,386),'max_seq_len':21},
"youtube":{'text_indices':(0,300),'audio_indices':(300,374),'video_indices':(374,409),'max_seq_len':21}
}
# use_context=True
# use_context_text=True
# use_context_audio=True
# use_context_video = True
# use_punchline_text=True
# use_punchline_audio=True
# use_punchline_video=True
experiment_configs=[
{'use_context':True,'use_punchline_text':True,'use_punchline_audio':True,'use_punchline_video':True,'use_context_text':True,'use_context_audio':True,'use_context_video':True},#ind 0:context,T+A+V
{'use_context':True,'use_punchline_text':True,'use_punchline_audio':False,'use_punchline_video':False,'use_context_text':True,'use_context_audio':False,'use_context_video':False},#ind 1:context,T
{'use_context':True,'use_punchline_text':True,'use_punchline_audio':True,'use_punchline_video':False,'use_context_text':True,'use_context_audio':True,'use_context_video':False},#ind 2:Context,T+A
{'use_context':True,'use_punchline_text':True,'use_punchline_audio':False,'use_punchline_video':True,'use_context_text':True,'use_context_audio':False,'use_context_video':True},#ind 3:context, T+V
{'use_context':False,'use_punchline_text':True,'use_punchline_audio':True,'use_punchline_video':True},#ind 4:No context,T+A+V
{'use_context':False,'use_punchline_text':True,'use_punchline_audio':False,'use_punchline_video':False},#ind 5: NO context, T
{'use_context':False,'use_punchline_text':True,'use_punchline_audio':True,'use_punchline_video':False},#ind 6: No context, T+A
{'use_context':False,'use_punchline_text':True,'use_punchline_audio':False,'use_punchline_video':True},#ind 7: No context, T+V
{'use_context':True,'use_punchline_text':False,'use_punchline_audio':True,'use_punchline_video':True,'use_context_text':False,'use_context_audio':True,'use_context_video':True},#ind:8,context,A+V
{'use_context':False,'use_punchline_text':False,'use_punchline_audio':True,'use_punchline_video':True},#ind:9,No Context,A+V
{'use_punchline':False,'use_context_text':True,'use_context_audio':True,'use_context_video':True},#ind:10,No punch,Context:T+A+V
{'use_punchline':False,'use_context_text':True,'use_context_audio':True,'use_context_video':False},#ind:11,No punch,Context:T+A
{'use_punchline':False,'use_context_text':True,'use_context_audio':False,'use_context_video':True},#ind:12,No punch,Context:T+V
{'use_punchline':False,'use_context_text':True,'use_context_audio':False,'use_context_video':False},#ind:13,No punch,Context:T
{'use_punchline':False,'use_context_text':False,'use_context_audio':True,'use_context_video':True}#ind:14,No punch,Context:A+V
]
num_experiments = len(experiment_configs)
#sacred will generate a different random _seed for every experiment
#and we will use that seed to control the randomness emanating from our libraries
node_index=30
#node_index=int(os.environ['SLURM_ARRAY_TASK_ID'])
#So, we are assuming that there will a folder called /processed_multimodal_data in the parent folder
#of this code. I wanted to keep it inside the .git folder. But git push limits file size to be <=100MB
#and some data files exceeds that size.
#all_datasets_location = "../processed_multimodal_data"
#due to limited space, we will directly use the data in mhoque lab
all_datasets_location = "/scratch/mhoque_lab/datasets/processed_multimodal_data/humor/CMFN"
two_context_t_a_v=1
selective_omission=2
discarding_punchline=3
emphaisis_on_a_subset=4
cur_experiment= emphaisis_on_a_subset
best_config = pickle.load(open("best_config_for_zooming.pkl","rb"))
def run_configs(dataset_location):
dataset_name = dataset_location[dataset_location.rfind("/")+1:]
if cur_experiment == two_context_t_a_v:
#print(best_config["unimodal_context"]["hidden_sizes"])
if node_index<2:
relevant_config = 0#to ramp up c+T+A+V
appropriate_config_dict = {**dataset_specific_config[dataset_name],**experiment_configs[relevant_config],"node_index":node_index,
"prototype":False,'dataset_location':dataset_location,"dataset_name":dataset_name,
"experiment_config_index":relevant_config,'epoch':80}
else:
appropriate_config_dict=best_config
hidden_text =random.choice([32,64,88,128,156,256,512])
hidden_audio = random.choice([8,16,32,48,64,80,90,100])
hidden_video = random.choice([8,16,32,48,64,80,90,100])
appropriate_config_dict["unimodal_context"]["hidden_sizes"] = [hidden_text,hidden_audio,hidden_video]
#print(appropriate_config_dict["unimodal_context"]["hidden_sizes"])
appropriate_config_dict["multimodal_context_configs"]['n_source_features'] = sum(appropriate_config_dict["unimodal_context"]["hidden_sizes"])
appropriate_config_dict["node_index"] = node_index
#appropriate_config_dict["prototype"]=True
#appropriate_config_dict["epoch"]=2
appropriate_config_dict.pop("device",None)
#print(appropriate_config_dict)
#print(appropriate_config_dict.keys())
r= ex.run(config_updates=appropriate_config_dict)
elif cur_experiment==discarding_punchline:
for relevant_config in range(10,15):
appropriate_config_dict = {**dataset_specific_config[dataset_name],**experiment_configs[relevant_config],"node_index":node_index,
"prototype":False,'dataset_location':dataset_location,"dataset_name":dataset_name,
"experiment_config_index":relevant_config}
r= ex.run(config_updates=appropriate_config_dict)
elif cur_experiment==emphaisis_on_a_subset:
for relevant_config in [0]:
appropriate_config_dict = {**dataset_specific_config[dataset_name],**experiment_configs[relevant_config],"node_index":node_index,
"prototype":True,'dataset_location':dataset_location,"dataset_name":dataset_name,
"experiment_config_index":relevant_config,"epoch":30}
r= ex.run(config_updates=appropriate_config_dict)
elif cur_experiment==selective_omission:
#for the selective feature omission, we are choosing context+T+A+V
relevant_config=0
num_omit_configs=17
while(True):
for omit_index in range(0,num_omit_configs):
appropriate_config_dict = {**dataset_specific_config[dataset_name],**experiment_configs[relevant_config],"node_index":node_index,
"prototype":False,'dataset_location':dataset_location,"dataset_name":dataset_name,
"experiment_config_index":relevant_config,"selectively_omitted_index":omit_index,"omit_corrected":"yes"}
r= ex.run(config_updates=appropriate_config_dict)
break
#print(appropriate_config_dict)
#Just run it many times
#r = ex.run(named_configs=['search_space'],config_updates={"node_index":node_index,"prototype":True})
#run it like ./running_different_configs.py --dataset=mosi
#or python running_different_configs.py --dataset=TED_humor
if __name__ == '__main__':
args = parser.parse_args()
dataset_path = os.path.join(all_datasets_location,args.dataset)
if(os.path.isdir(dataset_path)):
while(True):
run_configs(dataset_path)
else:
raise NotADirectoryError("Please input the dataset name correctly")