-
Notifications
You must be signed in to change notification settings - Fork 0
/
main_diff_eval_bfs.py
179 lines (149 loc) · 5.4 KB
/
main_diff_eval_bfs.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
import argparse
import pdb
import os
import sys
from torch.utils.data import DataLoader
import torch
from mimagen_pytorch import Unet3D, ElucidatedImagen, ImagenTrainer
from main_seq_bfs import Args as Args_seq
from main_diff_bfs import Args as Args_diff
sys.path.insert(0, './util')
from utils import read_args_txt
sys.path.insert(0, './data')
from data_bfs_preprocess import bfs_dataset
sys.path.insert(0, './transformer')
from sequentialModel import SequentialModel as transformer
sys.path.insert(0, './train_test_spatial')
from test_diff import test_final_overall
from test_diff_ensamble import test_final_overall_ensamble
class Args_final_eval:
def __init__(self):
self.parser = argparse.ArgumentParser()
"""
for finding the dynamics dir
"""
self.parser.add_argument("--bfs_dynamic_folder",
default='output/bfs_les_2023_12_21_12_09_10',
help='all the information of bfs training')
"""
reading the seq model
"""
self.parser.add_argument("--Nt_read",
default = 40,
help = "Which Nt model we need to read")
self.parser.add_argument("--use_best",
default = True)
"""
reading the diffusion model
"""
self.parser.add_argument("--Nepoch_read",
default = 2,
help = "Which epoch model we need to read")
"""
for dataset
"""
self.parser.add_argument("--trajec_max_len",
default=151,
help = 'max seq_length (per seq) to test the model')
self.parser.add_argument("--start_n",
default=9500, #9500 has problem
help = 'the starting step of the data')
self.parser.add_argument("--n_span",
default=152,
help='the total step of the data from the staring step')
"""
for seq_net_eval
"""
self.parser.add_argument("--test_Nt",
default=150,
help = 'How many step you want to proceed! Should be divided by 10')
"""
for eval dataset hyperparameter
"""
self.parser.add_argument("--batch_size", default = 1)
self.parser.add_argument("--device", type=str, default = "cuda:0")
def update_args(self):
args = self.parser.parse_args()
args.seq_args_txt = os.path.join(args.bfs_dynamic_folder,
'logging','args.txt' )
args.diff_args_txt = os.path.join(args.bfs_dynamic_folder,
'diffusion_folder',
'logging','args.txt')
# output dataset
args.experiment_path = os.path.join(args.bfs_dynamic_folder,
'diffusion_folder',
'experiment_final')
if not os.path.isdir(args.experiment_path):
os.makedirs(args.experiment_path)
return args
if __name__ == '__main__':
"""
Fetch args
"""
args_final = Args_final_eval()
args_final = args_final.update_args()
args_seq = read_args_txt(Args_seq(),
args_final.seq_args_txt)
args_diff = read_args_txt(Args_diff(),
args_final.diff_args_txt)
"""
Fetch dataset
"""
data_set = bfs_dataset(data_location = args_seq.data_location,
trajec_max_len = args_final.trajec_max_len,
start_n = args_final.start_n,
n_span = args_final.n_span)
data_loader = DataLoader(dataset=data_set,
shuffle=False,
batch_size=args_final.batch_size)
"""
Fetch models
"""
model = transformer(args_seq).to(args_final.device).float()
print('Number of parameters: {}'.format(model._num_parameters()))
if args_final.use_best:
model.load_state_dict(torch.load(args_seq.current_model_save_path+'best_model_sofar'))
else:
model.load_state_dict(torch.load(args_seq.current_model_save_path+'model_epoch_'+str(args_final.Nt_read),map_location=torch.device(args_final.device)))
unet1 = Unet3D(dim=args_diff.unet_dim,
cond_images_channels=2,
memory_efficient=True,
dim_mults=(1, 2, 4, 8)).to(torch.device(args_diff.device)) #mid: mid channel
image_sizes = (512)
image_width = (512)
imagen = ElucidatedImagen(
unets = (unet1),
image_sizes = image_sizes,
image_width = image_width,
channels = 2, # Han Gao add the input to this args explicity
random_crop_sizes = None,
num_sample_steps = args_diff.num_sample_steps, # original is 10
cond_drop_prob = 0.1,
sigma_min = 0.002,
sigma_max = (80), # max noise level, double the max noise level for upsampler （80，160）
sigma_data = 0.5, # standard deviation of data distribution
rho = 7, # controls the sampling schedule
P_mean = -1.2, # mean of log-normal distribution from which noise is drawn for training
P_std = 1.2, # standard deviation of log-normal distribution from which noise is drawn for training
S_churn = 80, # parameters for stochastic sampling - depends on dataset, Table 5 in apper
S_tmin = 0.05,
S_tmax = 50,
S_noise = 1.003,
condition_on_text = False,
auto_normalize_img = False # Han Gao make it false
).to(torch.device(args_final.device))
trainer = ImagenTrainer(imagen, device =torch.device(args_final.device))
trainer.load(path=args_diff.model_save_path+'/best_model_sofar')
test_final_overall_ensamble(args_final,
args_seq,
args_diff,
trainer,
model,
data_loader)
exit()
test_final_overall(args_final,
args_seq,
args_diff,
trainer,
model,
data_loader)