-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathtrain.py
187 lines (162 loc) · 7.7 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
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
import sys
import math
from typing import Iterable
import torch
import datasets as myDBs
from torchvision import transforms as vT
from util import audio_transforms as aT
from models.deepavfusion import DeepAVFusion
from models.avmae import AVMAE
from util import distributed as dist_utils
from util import misc as misc_utils
from util import data as data_utils
from util import meters, lr_sched
from util.knn_probe import EvalAVNNProbe
def main_worker(local_rank, args):
# Setup environment
job_dir = f"{args.output_dir}/{args.job_name}"
dist_utils.init_distributed_mode(local_rank, args, log_fn=f"{job_dir}/train.log")
device = torch.device('cpu') if not torch.cuda.is_available() else torch.device('cuda')
print(f'job dir: {job_dir}')
misc_utils.print_args(args)
# Adjust learning rate to batch size
num_tasks = dist_utils.get_world_size()
num_tasks_per_node = max(1, torch.cuda.device_count())
args.env.workers = args.env.workers // num_tasks_per_node
eff_batch_size = args.opt.batch_size * args.opt.accum_iter * num_tasks
if args.opt.lr is None: # only base_lr is specified
args.opt.lr = args.opt.blr * eff_batch_size / 256
print("base lr: %.2e" % args.opt.blr)
print("actual lr: %.2e" % args.opt.lr)
print("accumulate grad iterations: %d" % args.opt.accum_iter)
print("effective batch size: %d" % eff_batch_size)
# Dataloaders
dataset = myDBs.load_dataset(
args.data.dataset,
args.data.data_path,
dataset_type='simple',
visual_transform=vT.Compose([
vT.RandomResizedCrop(args.data.image_size, scale=(args.data.crop_min, 1.)),
vT.RandomHorizontalFlip(),
vT.ToTensor(),
vT.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])]),
audio_transform=aT.Compose([
aT.Pad(rate=args.data.audio_rate, dur=args.data.audio_dur),
aT.RandomVol(),
aT.MelSpectrogram(sample_rate=args.data.audio_rate, n_fft=int(args.data.audio_rate * 0.05), hop_length=int(args.data.audio_rate / 64), n_mels=args.data.audio_mels),
aT.Log()]),
train=True,
audio_dur=args.data.audio_dur,
audio_rate=args.data.audio_rate,
temporal_jitter=True,
)
loader = data_utils.get_dataloader(
dataset, args.env.distributed, args.opt.batch_size, args.env.workers, shuffle=True, drop_last=True)
print(dataset)
# Create model
image_size, audio_size = (args.data.image_size, args.data.image_size), (args.data.audio_mels, int(args.data.audio_dur * 64))
encoder = DeepAVFusion(
image_arch=args.model.image.backbone, image_pretrained=args.model.image.pretrained, image_size=image_size,
audio_arch=args.model.audio.backbone, audio_pretrained=args.model.audio.pretrained, audio_size=audio_size,
fusion_arch=args.model.fusion.arch,
fusion_layers=args.model.fusion.layers,
num_fusion_tkns=(args.model.fusion.num_fusion_tkns,
args.model.fusion.num_aggr_image_tkns,
args.model.fusion.num_aggr_audio_tkns),
fusion_mlp_ratio=args.model.fusion.mlp_ratio,
fusion_attn_ratio=args.model.fusion.attn_ratio,
fusion_num_heads=args.model.fusion.num_heads
)
model = AVMAE(
encoder, encoder.embed_dim,
image_decoder_arch=args.model.image.decoder_arch, image_decoder_depth=args.model.image.decoder_depth,
image_mask_ratio=args.model.image.mask_ratio, image_norm_loss=args.model.image.norm_loss,
audio_decoder_arch=args.model.audio.decoder_arch, audio_decoder_depth=args.model.audio.decoder_depth,
audio_mask_ratio=args.model.audio.mask_ratio, audio_norm_loss=args.model.audio.norm_loss
)
model.to(device)
print("Model = %s" % str(model))
# Optimizer
no_weight_decay_list = [n for n, p in model.named_parameters() if 'bias' in n or 'norm' in n]
param_groups = lr_sched.param_groups_pretrained(
model, args.opt.weight_decay, no_weight_decay_list=no_weight_decay_list,
image_pt=args.model.image.pretrained, audio_pt=args.model.audio.pretrained)
optimizer = torch.optim.AdamW(param_groups, lr=args.opt.lr, betas=(0.9, 0.95))
print(optimizer)
# Trainer
trainer = misc_utils.Trainer(
model,
optimizer=optimizer,
use_amp=args.opt.use_amp,
accum_iter=args.opt.accum_iter,
distributed=args.env.distributed
)
# Checkpointing and logging
ckpt_manager = misc_utils.CheckpointManager(
modules=trainer.module_dict(),
ckpt_dir=f"{job_dir}/checkpoints",
epochs=args.opt.epochs,
save_freq=args.log.save_freq)
start_epoch = ckpt_manager.resume()[0] if args.opt.resume else 0
wb_logger = misc_utils.WBLogger(
f"{job_dir}/wandb", args.log.wandb_entity, args.log.wandb_project, args.job_name,
model, args)
# Set up probes
knn_probe = EvalAVNNProbe(args.nn_probe, args.log, args.env)
# =============================================================== #
# Training loop
print(f"Start training for {args.opt.epochs} epochs")
for epoch in range(start_epoch, args.opt.epochs):
if args.env.distributed:
loader.sampler.set_epoch(epoch)
# train for one epoch
train_one_epoch(loader, trainer, epoch,
device=device, wb_logger=wb_logger, args=args)
# evaluate
if epoch % args.log.eval_freq == 0 or epoch == args.opt.epochs - 1 or epoch == start_epoch:
global_step = (len(loader) // trainer.accum_iter) * (epoch + 1)
knn_stats = knn_probe.evaluate(trainer.eval_model, epoch=epoch)
wb_logger.log(knn_stats, step=global_step, force=True)
# save checkpoint
ckpt_manager.checkpoint(epoch+1, {'epoch': epoch+1})
def train_one_epoch(loader: Iterable,
trainer: misc_utils.Trainer,
epoch: int = 0,
wb_logger: misc_utils.WBLogger = None,
device: torch.device = torch.device('cpu'),
args=None):
trainer.model.train(True)
metric_logger = meters.MetricLogger(delimiter=" ")
header = f'[Train][Ep-{epoch}/{args.opt.epochs}]'
trainer.zero_grad()
for step, (image, audio, _) in enumerate(metric_logger.log_every(loader, args.log.print_freq, header)):
sys.stdout.flush()
global_step = (len(loader) // trainer.accum_iter) * epoch + step // trainer.accum_iter
if step % args.opt.accum_iter == 0:
lr = lr_sched.adjust_learning_rate(trainer.optimizer, epoch + step / len(loader), args)
metric_logger.update(lr=lr)
# Prepare data
image = image.to(device, non_blocking=True).float()
audio = audio.to(device, non_blocking=True).float()
# Forward pass
with trainer.autocast(), trainer.autosync():
loss_image, loss_audio = trainer.model(image, audio)[:2]
loss = loss_image + loss_audio
if not math.isfinite(loss.item()):
raise f"Loss is {loss.item()}, stopping training"
# Backward pass and model update
grad_norm, amp_scale = trainer.step(loss)
# Log
if trainer.accums == 0:
metric_logger.update(
loss=loss.item(), loss_image=loss_image.item(), loss_audio=loss_audio.item(),
grad_norm=grad_norm, amp_scale=amp_scale, n=image.shape[0])
wb_logger.log(metric_logger.latest(), step=global_step)
if args.debug and step == 100:
break
# gather the stats from all processes
print("Syncing meters...")
metric_logger.synchronize_between_processes()
print("Averaged stats:", metric_logger)
trainer.zero_grad()
return metric_logger.averages()