-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmain.py
200 lines (175 loc) · 8.09 KB
/
main.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
import gc
import torch
import numpy as np
import random
import argparse
import os
import time
import warnings
warnings.simplefilter("ignore")
from algo.server.Real import Server as Real
from algo.server.Gen import Server as Gen
from algo.server.GenLLM import Server as GenLLM
from algo.server.Img2Cap import Server as Img2Cap
from algo.server.Filter import Server as Filter
from algo.server.Feedback import Server as Feedback
def run(args):
start = time.time()
timestamp = args.timestamp if args.timestamp else str(time.time())
print('Start timestamp:', timestamp)
i2i_strength = args.i2i_strength
for i in range(args.times):
print(f"\n============= Running time: {i}th =============")
print("Initiating ...")
args.i2i_strength = i2i_strength
args.task = os.path.join(
args.task_mode,
args.server_generator,
args.selector,
args.client_dataset,
timestamp,
str(i)
)
if args.framework == 'Real':
server = Real(args)
elif args.framework == 'Gen':
server = Gen(args)
elif args.framework == 'GenLLM':
server = GenLLM(args)
elif args.framework == 'Img2Cap':
server = Img2Cap(args)
elif args.framework == 'Filter':
server = Filter(args)
elif args.framework == 'Feedback':
server = Feedback(args)
else:
raise NotImplementedError
server.run()
del server
torch.cuda.empty_cache()
gc.collect()
print(f"\nTotal time cost: {round(time.time()-start, 2)}s.")
print("All done!")
if __name__ == "__main__":
total_start = time.time()
parser = argparse.ArgumentParser()
# general
parser.add_argument('-ts', "--timestamp", type=str, default="")
parser.add_argument('-tt', "--task_type", type=str, default="syn",
choices=[
"syn",
"mix"
])
parser.add_argument('-tm', "--task_mode", type=str, default="I2I",
choices=[
"T2I",
"I2I"
])
parser.add_argument('-ug', "--use_generated", type=bool, default=False)
parser.add_argument('-dev', "--device", type=str, default="cuda",
choices=[
"cpu",
"cuda"
])
parser.add_argument('-did', "--device_id", type=str, default="0")
parser.add_argument('-ab', "--auto_break", type=bool, default=False)
parser.add_argument('-tc', "--top_count", type=int, default=20,
help="For auto_break")
parser.add_argument('-T', "--times", type=int, default=1,
help="Running times")
parser.add_argument('-eg', "--eval_gap", type=int, default=1,
help="Rounds gap for evaluation")
parser.add_argument('-ddir', "--dataset_dir", type=str, default='./dataset',
help="A directory to save dataset")
parser.add_argument('-iter', "--iterations", type=int, default=100)
parser.add_argument('-eps', "--epsilon_per_iter", type=float, default=0.1,
help="Privacy budget per iteration")
parser.add_argument('-rvpl', "--real_volume_per_label", type=int, default=0)
parser.add_argument('-ims', "--image_max_size", type=int, default=256)
parser.add_argument('-vpl', "--volume_per_label", type=int, default=1)
parser.add_argument('-oa', "--online_api", type=bool, default=False)
parser.add_argument('-sgen', "--server_generator", type=str, default="StableDiffusion",
choices=[
"StableDiffusion",
"StableDiffusionXL",
"OpenJourney",
"FLUX",
])
parser.add_argument('-nipp', "--num_images_per_prompt", type=int, default=1)
parser.add_argument('-tr', "--test_ratio", type=float, default=0.2,
help="Used when the test set is not originally split")
parser.add_argument('-pml', "--prompt_max_length", type=int, default=77)
parser.add_argument('-f', "--framework", type=str, default="Gen",
choices=[
"Real",
"Gen",
"GenLLM",
"Img2Cap",
"Filter",
"Feedback"
])
parser.add_argument('-s', "--selector", type=str, default="Other")
parser.add_argument('-cdata', "--client_dataset", type=str, default="EuroSAT")
parser.add_argument('-cmodel', "--client_model", type=str, default="ResNet18",
help="CLIP, InceptionV3, ViTs, ResNets")
parser.add_argument('-cmp', "--client_model_pretrained", type=bool, default=False)
parser.add_argument('-cef', "--client_encoder_fixed", type=bool, default=False)
parser.add_argument('-cue', "--client_use_embedding", type=str, default="",
help="Refer to client_model")
parser.add_argument('-cret', "--client_retrain", type=bool, default=False)
parser.add_argument('-cbs', "--client_batch_size", type=int, default=16,
help="Edge clients require small batch size")
parser.add_argument('-clr', "--client_learning_rate", type=float, default=0.001)
parser.add_argument('-ce', "--client_epochs", type=int, default=100)
parser.add_argument('-cuf', "--client_use_filtered", type=bool, default=False)
parser.add_argument('-caf', "--client_accumulate_filter", type=bool, default=False)
parser.add_argument('-cst', "--client_send_topk", type=bool, default=False)
parser.add_argument('-ctpl', "--client_topk_per_label", type=int, default=10000)
# I2I
parser.add_argument('-is', "--i2i_strength", type=float, default=0.8,
help="[0,1]")
parser.add_argument('-isa', "--i2i_strength_anneal", type=float, default=0.02)
parser.add_argument('-isth', "--i2i_strength_threshold", type=float, default=0.6)
parser.add_argument('-uipa', "--use_IPAdapter", type=bool, default=False)
parser.add_argument('-ipas', "--IPAdapter_scale", type=float, default=0.2,
help="[0,1]")
# GenLLM
parser.add_argument('-sllm', "--server_llm", type=str, default="",
choices=[
"",
"Llama2",
"Llama3"
])
# Img2Cap
parser.add_argument('-cml', "--caption_max_length", type=int, default=70)
parser.add_argument('-scap', "--server_captioner", type=str, default="",
choices=[
"",
"BlipBase",
"BlipLarge",
"LLaVA"
])
# RF
parser.add_argument('-dth', "--dist_threshold", type=float, default=0.0)
# PE
parser.add_argument('-hth', "--histogram_threshold", type=int, default=0)
args = parser.parse_args()
os.environ["CUDA_VISIBLE_DEVICES"] = args.device_id
if args.use_generated:
assert args.timestamp, 'timestamp is required when use_generated=True'
if args.device == "cuda" and not torch.cuda.is_available():
print("\ncuda is not avaiable.\n")
args.device = "cpu"
if args.task_type == "mix":
assert args.real_volume_per_label > 0, 'real_volume_per_label should > 0 when task_type == "mix"'
if args.framework not in ["Filter", "Feedback"]:
args.iterations = 1
elif args.framework == "Filter" and not args.client_accumulate_filter:
args.iterations = 1
if args.framework == "Filter":
assert args.client_use_filtered, 'client_use_filtered should be True when framework == "Filter"'
print("=" * 50)
for arg in vars(args):
print(arg, '=',getattr(args, arg))
print("=" * 50)
run(args)