forked from jraine/flows4flows
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathrun_jointbase.py
166 lines (139 loc) · 7.34 KB
/
run_jointbase.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
import hydra
from omegaconf import DictConfig, OmegaConf
import pathlib
import glob
import os
from ffflows.models import BaseFlow
from ffflows.utils import set_trainable
import torch
from torch.utils.data import DataLoader
from nflows.distributions import StandardNormal
from utils import get_activation, get_flow4flow, train, spline_inn, get_conditional_data, tensor_to_str, \
set_penalty, get_flow4flow_ncond, dump_to_df
from plot import plot_data, plot_arrays
from ffflows.data.dist_to_dist import ConditionalDataToData, ConditionalDataToTarget
import numpy as np
np.random.seed(42)
torch.manual_seed(42)
def train_base(*args, **kwargs):
return train(*args, **kwargs, rand_perm_target=False)
def train_f4f(*args, **kwargs):
return train(*args, **kwargs, rand_perm_target=True)
@hydra.main(version_base=None, config_path="conf/", config_name="defaults_jointcond")
def main(cfg: DictConfig) -> None:
print("Configuring job with following options")
print(OmegaConf.to_yaml(cfg))
outputpath = pathlib.Path(cfg.output.save_dir + '/' + cfg.output.name)
outputpath.mkdir(parents=True, exist_ok=True)
with open(outputpath / f"{cfg.output.name}.yaml", 'w') as file:
OmegaConf.save(config=cfg, f=file)
if cfg.general.ncond is None or cfg.general.ncond < 1:
print(
f"Cannot train Flows4Flows on the same base distribution without any conditions. You specified cfg.general.ncond = {cfg.general.ncond}. Exiting now.")
exit(42)
# Set device
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
# Get training data
def get_data(n_points):
return get_conditional_data(cfg.base_dist.condition, cfg.base_dist.data, n_points)
n_points = int(cfg.general.npoints)
base_data = DataLoader(
dataset=get_data(n_points),
batch_size=cfg.base_dist.batch_size,
shuffle=True
)
val_base_data = DataLoader(
dataset=get_data(n_points),
batch_size=1000
)
# Train base
base_flow = BaseFlow(spline_inn(cfg.general.data_dim,
nodes=cfg.base_dist.nnodes,
num_blocks=cfg.base_dist.nblocks,
num_stack=cfg.base_dist.nstack,
tail_bound=4.0,
activation=get_activation(cfg.base_dist.activation),
num_bins=cfg.base_dist.nbins,
context_features=cfg.general.ncond
),
StandardNormal([cfg.general.data_dim])
)
if pathlib.Path(cfg.base_dist.load_path).is_file():
print(f"Loading base from model: {cfg.base_dist.load_path}")
base_flow.load_state_dict(torch.load(cfg.base_dist.load_path, map_location=device))
else:
print("Training base distribution")
train_base(base_flow, base_data, val_base_data,
cfg.base_dist.nepochs, cfg.base_dist.lr, cfg.general.ncond,
outputpath, name='base', device=device, gclip=cfg.base_dist.gclip)
with open(outputpath / f'base' / f'{cfg.base_dist.condition[:3].lower()}_{cfg.base_dist.data.lower()}.yaml', 'w') as file:
models = glob.glob(str((outputpath / f'base' / 'epoch*pt').resolve()))
models.sort(key=os.path.getmtime)
cfg.base_dist.load_path = models[-1]
OmegaConf.save(config=cfg.base_dist, f=file)
set_trainable(base_flow, False)
base_flow.to(device)
nevalpoints = 6
bd_samples = []
with torch.no_grad():
for right_cond in (evals := get_data(20).get_default_eval(nevalpoints)):
nsamples = int(1e5)
right_cond = torch.Tensor([right_cond]).view(1, -1).to(device)
plot_data(sampled := base_flow.sample(nsamples, context=right_cond, batch_size=int(1e5)).view(-1, 2),
outputpath / f'base_density_{tensor_to_str(right_cond)}.png')
bd_samples.append(sampled)
df = dump_to_df(*bd_samples,
col_names=[f'cond_{ev:.2f}_{coord}'.replace('.', '_') for ev in evals for coord in ['x', 'y']])
df.to_hdf(outputpath / 'eval_data.h5', f'base_dist')
# Train Flow4Flow
n_cond = get_flow4flow_ncond(cfg.top_transformer.flow4flow) * cfg.general.ncond
f4flow = get_flow4flow(cfg.top_transformer.flow4flow,
spline_inn(cfg.general.data_dim,
nodes=cfg.top_transformer.nnodes,
num_blocks=cfg.top_transformer.nblocks,
num_stack=cfg.top_transformer.nstack,
tail_bound=4.0,
activation=get_activation(cfg.top_transformer.activation),
num_bins=cfg.top_transformer.nbins,
context_features=n_cond,
flow_for_flow=True
),
base_flow)
set_penalty(f4flow, cfg.top_transformer.penalty, cfg.top_transformer.penalty_weight, cfg.top_transformer.anneal)
if pathlib.Path(cfg.top_transformer.load_path).is_file():
print(f"Loading Flow4Flow from model: {cfg.top_transformer.load_path}")
f4flow.load_state_dict(torch.load(cfg.top_transformer.load_path, map_location=device))
else:
print("Training Flow4Flow model")
train_f4f(f4flow, base_data, val_base_data,
cfg.top_transformer.nepochs, cfg.top_transformer.lr, cfg.general.ncond,
outputpath, name='f4f', device=device, gclip=cfg.top_transformer.gclip)
with torch.no_grad():
f4flow.to(device)
test_data = get_data(n_points)
test_points = test_data.get_default_eval(6)
for con in test_points:
# Handle the broadcasting
left_data, left_cond, right_cond = [d.to(device) \
for d in ConditionalDataToTarget(test_data.get_tuple(), con).paired()]
# Transform the data
transformed, _ = f4flow.batch_transform(left_data, left_cond, right_cond, batch_size=1000)
# Plot the output densities
plot_data(transformed, outputpath / f'flow_for_flow_{tensor_to_str(con)}.png')
# Get the transformation that results from going via the base distributions
left_bd_enc = f4flow.base_flow_left.transform_to_noise(left_data, left_cond)
right_bd_dec, _ = f4flow.base_flow_right._transform.inverse(left_bd_enc, right_cond)
# Plot how each point is shifted
plot_arrays({
'Input Data': left_data,
'FFF': transformed,
'BdTransfer': right_bd_dec
}, outputpath, f'{con.item():.2f}')
##dump data
df = dump_to_df(left_data, left_cond, right_cond, transformed, left_bd_enc, right_bd_dec,
col_names=['input_x','input_y','left_cond','right_cond',
'transformed_x','transformed_y','left_enc_x','left_enc_y',
'base_transfer_x','base_transfer_y'])
df.to_hdf(outputpath / 'eval_data.h5', f'f4f_{con:.2f}'.replace('.','_'))
if __name__ == "__main__":
main()