-
Notifications
You must be signed in to change notification settings - Fork 5
/
train_ENARF_GAN.py
311 lines (244 loc) · 12.1 KB
/
train_ENARF_GAN.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
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
import argparse
import os
import time
import warnings
import tensorboardX as tbx
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import DataLoader
from dataset.dataset import HumanDataset, HumanPoseDataset
from libraries.config import yaml_config
from libraries.custom_stylegan2.net import Discriminator
from libraries.gan.loss import adv_loss_dis, adv_loss_gen, d_r1_loss
from libraries.train_utils import record_command, write, save_img
from models.generator import TriNARFGenerator
from models.loss import nerf_patch_loss
warnings.filterwarnings('ignore')
def train(train_func, config):
datasets, data_loaders = create_dataloader(config.dataset)
train_func(config, datasets, data_loaders, rank=0, ddp=False)
def cache_dataset(config_dataset):
create_dataset(config_dataset, just_cache=True)
def create_dataset(config_dataset, just_cache=False):
size = config_dataset.image_size
dataset_name = config_dataset.name
train_dataset_config = config_dataset.train
print("loading datasets")
if dataset_name == "human_v2":
img_dataset = HumanDataset(train_dataset_config, size=size, return_bone_params=False,
just_cache=just_cache)
pose_prior_root = train_dataset_config.pose_prior_root or train_dataset_config.data_root
print("pose prior:", pose_prior_root)
pose_dataset = HumanPoseDataset(size=size, data_root=pose_prior_root,
just_cache=just_cache)
else:
assert False
return img_dataset, pose_dataset
def create_dataloader(config_dataset):
batch_size = config_dataset.bs
shuffle = True
drop_last = True
num_workers = config_dataset.num_workers
print("num_workers:", num_workers)
img_dataset, pose_dataset = create_dataset(config_dataset)
loader_img = DataLoader(img_dataset, batch_size=batch_size, num_workers=num_workers, shuffle=shuffle,
drop_last=drop_last)
loader_pose = DataLoader(pose_dataset, batch_size=batch_size, num_workers=num_workers, shuffle=shuffle,
drop_last=drop_last)
return (img_dataset, pose_dataset), (loader_img, loader_pose)
def prepare_models(gen_config, dis_config, pose_dataset, size):
if gen_config.use_triplane:
gen = TriNARFGenerator(gen_config, size, num_bone=pose_dataset.num_bone,
num_bone_param=pose_dataset.num_bone_param,
parent_id=pose_dataset.parents)
gen.register_canonical_pose(pose_dataset.canonical_pose)
else:
raise ValueError("generator without triplane is not supported")
dis = Discriminator(dis_config, size=size)
return gen, dis
def loss(gen, dis, fake_img, fake_low_res_mask, bone_mask, background_ratio,
bone_loss_func, gen_optimizer, dis_optimizer, adv_loss_type, ddp,
world_size):
loss_dict = {}
loss_bone = bone_loss_func(fake_low_res_mask, bone_mask,
background_ratio) * config.loss.bone_guided_coef
dis_fake = dis(fake_img, ddp, world_size)
gen_optimizer.zero_grad(set_to_none=True)
dis_optimizer.zero_grad(set_to_none=True)
loss_adv_gen = adv_loss_gen(dis_fake, adv_loss_type, tmp=1)
loss_gen = loss_adv_gen + loss_bone
if config.loss.tri_plane_reg_coef > 0:
loss_triplane = gen.nerf.buffers_tensors["tri_plane_feature"].square().mean()
loss_gen += loss_triplane * config.loss.tri_plane_reg_coef
loss_dict["adv_loss_gen"] = loss_adv_gen
loss_dict["bone_loss"] = loss_bone
return loss_gen, loss_dict
def train_step(iter, batchsize, gen, pose_to_camera, pose_to_world, bone_length, inv_intrinsic,
bone_loss_func, bone_mask, dis, ddp, world_size, gen_optimizer, dis_optimizer,
adv_loss_type, rank, writer, real_img, r1_loss_coef):
n_accum_step = config.n_accum_step
forward_bs = batchsize // n_accum_step
fake_img = []
dis.requires_grad_(False)
for i in range(0, batchsize, forward_bs):
# randomly sample latent
z = torch.cuda.FloatTensor(forward_bs, config.generator_params.z_dim * 4).normal_()
fake_img_i, fake_mask_i, fine_weights, fine_depth = gen(pose_to_camera[i:i + forward_bs],
pose_to_world[i:i + forward_bs],
bone_length[i:i + forward_bs], z,
inv_intrinsic[i:i + forward_bs])
background_ratio = gen.background_ratio
loss_gen, loss_dict = loss(gen, dis, fake_img_i, fake_mask_i,
bone_mask[i:i + forward_bs], background_ratio, bone_loss_func, gen_optimizer,
dis_optimizer, adv_loss_type, ddp, world_size)
loss_gen.backward()
fake_img.append(fake_img_i)
fake_img = torch.cat(fake_img)
gen_optimizer.step()
if rank == 0:
if iter % 100 == 0:
print(iter)
for k, v in loss_dict.items():
write(iter, v, k, writer)
torch.cuda.empty_cache()
# update discriminator
gen_optimizer.zero_grad(set_to_none=True)
dis_optimizer.zero_grad(set_to_none=True)
dis.requires_grad_(True)
dis_fake = dis(fake_img.detach(), ddp, world_size)
dis_real = dis(real_img, ddp, world_size)
loss_dis = adv_loss_dis(dis_real, dis_fake, adv_loss_type)
if rank == 0:
if iter % 100 == 0:
write(iter, loss_dis, "adv_loss_dis", writer)
loss_dis.backward()
dis_optimizer.step()
if iter % 16 == 0:
gen_optimizer.zero_grad(set_to_none=True)
dis_optimizer.zero_grad(set_to_none=True)
real_img.requires_grad = True
torch.cuda.empty_cache()
dis_real = dis(real_img, ddp, world_size)
r1_loss = d_r1_loss(dis_real, real_img)
if rank == 0:
write(iter, r1_loss, "r1_reg", writer)
(1 / 2 * r1_loss * 16 * r1_loss_coef + 0 * dis_real[
0]).backward() # 0 * dis_real[0] avoids zero grad
dis_optimizer.step()
torch.cuda.empty_cache()
return fake_img
def train_func(config, datasets, data_loaders, rank, ddp=False, world_size=1):
torch.backends.cudnn.benchmark = True
out_dir = config.out_root
out_name = config.out
if rank == 0:
writer = tbx.SummaryWriter(f"{out_dir}/runs/{out_name}")
os.makedirs(f"{out_dir}/result/{out_name}", exist_ok=True)
record_command(f"{out_dir}/result/{out_name}")
else:
writer = None
size = config.dataset.image_size
num_iter = config.num_iter
batchsize = config.dataset.bs
adv_loss_type = config.loss.adv_loss_type
r1_loss_coef = config.loss.r1_loss_coef
img_dataset, pose_dataset = datasets
loader_img, loader_pose = data_loaders
gen, dis = prepare_models(config.generator_params, config.discriminator_params, pose_dataset, size)
num_gpus = torch.cuda.device_count()
n_gpu = rank % num_gpus
torch.cuda.set_device(n_gpu)
gen = gen.cuda(n_gpu)
dis = dis.cuda(n_gpu)
if ddp:
gen = torch.nn.SyncBatchNorm.convert_sync_batchnorm(gen)
gen = nn.parallel.DistributedDataParallel(gen, device_ids=[n_gpu])
dis = nn.parallel.DistributedDataParallel(dis, device_ids=[n_gpu])
bone_loss_func = nerf_patch_loss
gen_lr = 1e-3 * batchsize / 32
dis_lr = 2e-3 * batchsize / 32
gen_optimizer = optim.Adam(gen.parameters(), lr=gen_lr, betas=(0, 0.99))
dis_optimizer = optim.Adam(dis.parameters(), lr=dis_lr, betas=(0, 0.99))
iter = 0
start_time = time.time()
if config.resume or config.resume_latest:
path = f"{out_dir}/result/{out_name}/snapshot_latest.pth" if config.resume_latest else config.resume
if os.path.exists(path):
snapshot = torch.load(path, map_location="cuda")
if ddp:
gen_module = gen.module
dis_module = dis.module
else:
gen_module = gen
dis_module = dis
gen_module.load_state_dict(snapshot["gen"], strict=False)
dis_module.load_state_dict(snapshot["dis"])
# gen_optimizer.load_state_dict(snapshot["gen_opt"])
# dis_optimizer.load_state_dict(snapshot["dis_opt"])
iter = snapshot["iteration"]
del snapshot
init_iter = iter
while iter < num_iter:
for i, (img, pose) in enumerate(zip(loader_img, loader_pose)):
if (iter + 1) % 10 == 0 and rank == 0:
print(f"{iter + 1} iter, {(time.time() - start_time) / (iter - init_iter + 1)} s/iter")
gen.train()
dis.train()
dis.requires_grad_(False)
real_img = img["img"].cuda(non_blocking=True).float()
bone_mask = pose["bone_mask"].cuda(non_blocking=True)
pose_to_camera = pose["pose_to_camera"].cuda(non_blocking=True)
bone_length = pose["bone_length"].cuda(non_blocking=True)
pose_to_world = pose["pose_to_world"].cuda(non_blocking=True)
intrinsic = pose["intrinsics"].cuda(non_blocking=True)
inv_intrinsic = torch.inverse(intrinsic)
if real_img.shape[0] != batchsize or bone_mask.shape[0] != batchsize: # drop last minibatch
continue
# fake_img = train_step(iter, batchsize, gen, pose_to_camera, pose_to_world, bone_length, inv_intrinsic,
# bone_loss_func, bone_mask, dis, ddp, world_size, gen_optimizer, dis_optimizer,
# adv_loss_type, rank, writer, real_img, r1_loss_coef)
try:
fake_img = train_step(iter, batchsize, gen, pose_to_camera, pose_to_world, bone_length, inv_intrinsic,
bone_loss_func, bone_mask, dis, ddp, world_size, gen_optimizer, dis_optimizer,
adv_loss_type, rank, writer, real_img, r1_loss_coef)
except:
print("iteration skipped")
torch.cuda.empty_cache()
continue
if rank == 0:
if iter == 10:
with open(f"{out_dir}/result/{out_name}/iter_10_succeeded.txt", "w") as f:
f.write("ok")
if iter % 50 == 0:
save_img(fake_img, f"{out_dir}/result/{out_name}/rgb_{iter // 5000 * 5000}.png")
save_img(real_img, f"{out_dir}/result/{out_name}/real.png")
save_img(bone_mask, f"{out_dir}/result/{out_name}/bone_{iter // 5000 * 5000}.png")
if (iter + 1) % 200 == 0:
if ddp:
gen_module = gen.module
dis_module = dis.module
else:
gen_module = gen
dis_module = dis
save_params = {"iteration": iter,
"start_time": start_time,
"gen": gen_module.state_dict(),
"dis": dis_module.state_dict(),
"gen_opt": gen_optimizer.state_dict(),
"dis_opt": dis_optimizer.state_dict(),
}
torch.save(save_params, f"{out_dir}/result/{out_name}/snapshot_latest.pth")
torch.save(save_params,
f"{out_dir}/result/{out_name}/snapshot_{(iter // 50000 + 1) * 50000}.pth")
torch.cuda.empty_cache()
iter += 1
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument('--config', type=str, default="configs/enarfgan/AIST/enarfgan.yml")
parser.add_argument('--default_config', type=str, default="configs/enarfgan/default.yml")
parser.add_argument('--resume_latest', action="store_true")
parser.add_argument('--num_workers', type=int, default=1)
args = parser.parse_args()
config = yaml_config(args.config, args.default_config, args.resume_latest, args.num_workers)
train(train_func, config)