-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrain.py
169 lines (150 loc) · 8.3 KB
/
train.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
import os
import cv2
import math
import time
import torch
import torch.distributed as dist
import numpy as np
import random
import argparse
from model.RIFE import Model
from dataset import *
from torch.utils.data import DataLoader, Dataset
from torch.utils.tensorboard import SummaryWriter
from torch.utils.data.distributed import DistributedSampler
device = torch.device("cuda")
log_path = 'train_log'
def get_learning_rate(step):
if step < 2000:
mul = step / 2000.
return 3e-4 * mul
else:
mul = np.cos((step - 2000) / (args.epoch * args.step_per_epoch - 2000.) * math.pi) * 0.5 + 0.5
return (3e-4 - 3e-6) * mul + 3e-6
def flow2rgb(flow_map_np): #将一个光流向量场转化为一种可以直观可视化的RGB格式
h, w, _ = flow_map_np.shape #获取flow_map_np的形状，高度h,宽度w,颜色通道数不使用
rgb_map = np.ones((h, w, 3)).astype(np.float32) #创建一个全1的数组，用于存储转换后的RGB图像
normalized_flow_map = flow_map_np / (np.abs(flow_map_np).max()) #】归一化光流图的值
#normalized_flow_map是归一化的光流图
rgb_map[:, :, 0] += normalized_flow_map[:, :, 0]
rgb_map[:, :, 1] -= 0.5 * (normalized_flow_map[:, :, 0] + normalized_flow_map[:, :, 1])
rgb_map[:, :, 2] += normalized_flow_map[:, :, 1]
return rgb_map.clip(0, 1)
def train(model, local_rank): #model为训练模型，local_rank为分布式训练的本地排名
if local_rank == 0: #根据local_rank的值初始化了TensorBoard的日志记录器，SummaryWritter,若local_rank为0,通常表示主节点，则在train和validate目录下创建日志记录器，否则不记录
writer = SummaryWriter('train')
writer_val = SummaryWriter('validate')
else:
writer = None
writer_val = None
step = 0 #训练步骤计数器
nr_eval = 0 #评估计数器
dataset = VimeoDataset('train') #创建一个dataset vimeo
sampler = DistributedSampler(dataset) #采用分布式采样器DistributedSample
train_data = DataLoader(dataset, batch_size=args.batch_size, num_workers=8, pin_memory=True, drop_last=True, sampler=sampler)
#num_workers=8 8个进程的处理数据
#pin_memory=True Dataloader 返回之前，将数据张量复制到CUDA固定内存之中，有利于数据更快的转移到GPU中
#drop_last=True 丢弃最后一个数据集（数量不满一个batch）
#sample=sample,表示从数据集中抽取书数据的策略，这里指DistributedSample
args.step_per_epoch = train_data.__len__() #计算每个epoch的步数并且保存到args.step_per_epoch
#加载一个验证dataset_val和val_data
dataset_val = VimeoDataset('validation')
val_data = DataLoader(dataset_val, batch_size=16, pin_memory=True, num_workers=8)
print('training...')
time_stamp = time.time() #记录时间戳
for epoch in range(args.epoch):
sampler.set_epoch(epoch) #设置epoch以确保数据的随机性
for i, data in enumerate(train_data):
data_time_interval = time.time() - time_stamp
time_stamp = time.time()
data_gpu, timestep = data
data_gpu = data_gpu.to(device, non_blocking=True) / 255.
timestep = timestep.to(device, non_blocking=True)
imgs = data_gpu[:, :6]
gt = data_gpu[:, 6:9]
learning_rate = get_learning_rate(step) * args.world_size / 4
pred, info = model.update(imgs, gt, learning_rate, training=True) # pass timestep if you are training RIFEm
# 使用模型进行一次训练迭代，获取预测结果和其他信息
train_time_interval = time.time() - time_stamp
time_stamp = time.time()
if step % 200 == 1 and local_rank == 0:
writer.add_scalar('learning_rate', learning_rate, step)
writer.add_scalar('loss/l1', info['loss_l1'], step)
writer.add_scalar('loss/tea', info['loss_tea'], step)
writer.add_scalar('loss/distill', info['loss_distill'], step)
if step % 1000 == 1 and local_rank == 0:
gt = (gt.permute(0, 2, 3, 1).detach().cpu().numpy() * 255).astype('uint8')
mask = (torch.cat((info['mask'], info['mask_tea']), 3).permute(0, 2, 3, 1).detach().cpu().numpy() * 255).astype('uint8')
pred = (pred.permute(0, 2, 3, 1).detach().cpu().numpy() * 255).astype('uint8')
merged_img = (info['merged_tea'].permute(0, 2, 3, 1).detach().cpu().numpy() * 255).astype('uint8')
flow0 = info['flow'].permute(0, 2, 3, 1).detach().cpu().numpy()
flow1 = info['flow_tea'].permute(0, 2, 3, 1).detach().cpu().numpy()
for i in range(5):
imgs = np.concatenate((merged_img[i], pred[i], gt[i]), 1)[:, :, ::-1]
writer.add_image(str(i) + '/img', imgs, step, dataformats='HWC')
writer.add_image(str(i) + '/flow', np.concatenate((flow2rgb(flow0[i]), flow2rgb(flow1[i])), 1), step, dataformats='HWC')
writer.add_image(str(i) + '/mask', mask[i], step, dataformats='HWC')
writer.flush()
if local_rank == 0:
print('epoch:{} {}/{} time:{:.2f}+{:.2f} loss_l1:{:.4e}'.format(epoch, i, args.step_per_epoch, data_time_interval, train_time_interval, info['loss_l1']))
step += 1
nr_eval += 1
if nr_eval % 5 == 0:
evaluate(model, val_data, step, local_rank, writer_val)
model.save_model(log_path, local_rank)
dist.barrier()
def evaluate(model, val_data, nr_eval, local_rank, writer_val):
loss_l1_list = []
loss_distill_list = []
loss_tea_list = []
psnr_list = []
psnr_list_teacher = []
time_stamp = time.time()
for i, data in enumerate(val_data):
data_gpu, timestep = data
data_gpu = data_gpu.to(device, non_blocking=True) / 255.
imgs = data_gpu[:, :6]
gt = data_gpu[:, 6:9]
with torch.no_grad():
pred, info = model.update(imgs, gt, training=False)
merged_img = info['merged_tea']
loss_l1_list.append(info['loss_l1'].cpu().numpy())
loss_tea_list.append(info['loss_tea'].cpu().numpy())
loss_distill_list.append(info['loss_distill'].cpu().numpy())
for j in range(gt.shape[0]):
psnr = -10 * math.log10(torch.mean((gt[j] - pred[j]) * (gt[j] - pred[j])).cpu().data)
psnr_list.append(psnr)
psnr = -10 * math.log10(torch.mean((merged_img[j] - gt[j]) * (merged_img[j] - gt[j])).cpu().data)
psnr_list_teacher.append(psnr)
gt = (gt.permute(0, 2, 3, 1).cpu().numpy() * 255).astype('uint8')
pred = (pred.permute(0, 2, 3, 1).cpu().numpy() * 255).astype('uint8')
merged_img = (merged_img.permute(0, 2, 3, 1).cpu().numpy() * 255).astype('uint8')
flow0 = info['flow'].permute(0, 2, 3, 1).cpu().numpy()
flow1 = info['flow_tea'].permute(0, 2, 3, 1).cpu().numpy()
if i == 0 and local_rank == 0:
for j in range(10):
imgs = np.concatenate((merged_img[j], pred[j], gt[j]), 1)[:, :, ::-1]
writer_val.add_image(str(j) + '/img', imgs.copy(), nr_eval, dataformats='HWC')
writer_val.add_image(str(j) + '/flow', flow2rgb(flow0[j][:, :, ::-1]), nr_eval, dataformats='HWC')
eval_time_interval = time.time() - time_stamp
if local_rank != 0:
return
writer_val.add_scalar('psnr', np.array(psnr_list).mean(), nr_eval)
writer_val.add_scalar('psnr_teacher', np.array(psnr_list_teacher).mean(), nr_eval)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument('--epoch', default=200, type=int)
parser.add_argument('--batch_size', default=16, type=int, help='minibatch size')
parser.add_argument('--local_rank', default=0, type=int, help='local rank')
parser.add_argument('--world_size', default=4, type=int, help='world size')
args = parser.parse_args()
torch.distributed.init_process_group(backend="nccl", world_size=args.world_size)
torch.cuda.set_device(args.local_rank)
seed = 1234
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
torch.backends.cudnn.benchmark = True
model = Model(args.local_rank)
train(model, args.local_rank)