zgcr654321
commited on
Commit
•
a06b5ad
1
Parent(s):
c1ca0dc
Upload 28 files
Browse files- imagenet/van_b0/__pycache__/train_config.cpython-38.pyc +0 -0
- imagenet/van_b0/checkpoints/latest.pth +3 -0
- imagenet/van_b0/checkpoints/van_b0-acc75.618.pth +3 -0
- imagenet/van_b0/log/train.info.log +0 -0
- imagenet/van_b0/log/train.info.log.2023-11-21 +0 -0
- imagenet/van_b0/log/train.info.log.2023-11-28 +0 -0
- imagenet/van_b0/test.sh +1 -0
- imagenet/van_b0/test_config.py +55 -0
- imagenet/van_b0/train.sh +1 -0
- imagenet/van_b0/train_config.py +126 -0
- imagenet/van_b1/__pycache__/train_config.cpython-38.pyc +0 -0
- imagenet/van_b1/checkpoints/latest.pth +3 -0
- imagenet/van_b1/checkpoints/van_b1-acc80.956.pth +3 -0
- imagenet/van_b1/log/train.info.log +0 -0
- imagenet/van_b1/log/train.info.log.2023-11-28 +0 -0
- imagenet/van_b1/test.sh +1 -0
- imagenet/van_b1/test_config.py +55 -0
- imagenet/van_b1/train.sh +1 -0
- imagenet/van_b1/train_config.py +126 -0
- imagenet/van_b2/__pycache__/train_config.cpython-38.pyc +0 -0
- imagenet/van_b2/checkpoints/latest.pth +3 -0
- imagenet/van_b2/checkpoints/van_b2-acc82.322.pth +3 -0
- imagenet/van_b2/log/train.info.log +0 -0
- imagenet/van_b2/log/train.info.log.2023-11-21 +0 -0
- imagenet/van_b2/test.sh +1 -0
- imagenet/van_b2/test_config.py +55 -0
- imagenet/van_b2/train.sh +1 -0
- imagenet/van_b2/train_config.py +126 -0
imagenet/van_b0/__pycache__/train_config.cpython-38.pyc
ADDED
Binary file (3.15 kB). View file
|
|
imagenet/van_b0/checkpoints/latest.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5e3973460b358593771c4a8c864bb3033f55034f96d4443dd26b71c653ab986c
|
3 |
+
size 49711051
|
imagenet/van_b0/checkpoints/van_b0-acc75.618.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:162b506f60e4a1f1251c9e7de97bd9324ccf758474727277531cd27ee6bce45b
|
3 |
+
size 16575333
|
imagenet/van_b0/log/train.info.log
ADDED
The diff for this file is too large to render.
See raw diff
|
|
imagenet/van_b0/log/train.info.log.2023-11-21
ADDED
The diff for this file is too large to render.
See raw diff
|
|
imagenet/van_b0/log/train.info.log.2023-11-28
ADDED
The diff for this file is too large to render.
See raw diff
|
|
imagenet/van_b0/test.sh
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.run --nproc_per_node=2 --master_addr 127.0.1.0 --master_port 10000 ../../../tools/test_classification_model.py --work-dir ./
|
imagenet/van_b0/test_config.py
ADDED
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import sys
|
3 |
+
|
4 |
+
BASE_DIR = os.path.dirname(
|
5 |
+
os.path.dirname(os.path.dirname(os.path.dirname(
|
6 |
+
os.path.abspath(__file__)))))
|
7 |
+
sys.path.append(BASE_DIR)
|
8 |
+
|
9 |
+
from tools.path import ILSVRC2012_path
|
10 |
+
|
11 |
+
from simpleAICV.classification import backbones
|
12 |
+
from simpleAICV.classification import losses
|
13 |
+
from simpleAICV.classification.datasets.ilsvrc2012dataset import ILSVRC2012Dataset
|
14 |
+
from simpleAICV.classification.common import Opencv2PIL, TorchResize, TorchCenterCrop, TorchMeanStdNormalize, ClassificationCollater, load_state_dict
|
15 |
+
|
16 |
+
import torch
|
17 |
+
import torchvision.transforms as transforms
|
18 |
+
|
19 |
+
|
20 |
+
class config:
|
21 |
+
'''
|
22 |
+
for resnet,input_image_size = 224;for darknet,input_image_size = 256
|
23 |
+
'''
|
24 |
+
network = 'van_b0'
|
25 |
+
num_classes = 1000
|
26 |
+
input_image_size = 224
|
27 |
+
scale = 256 / 224
|
28 |
+
|
29 |
+
model = backbones.__dict__[network](**{
|
30 |
+
'num_classes': num_classes,
|
31 |
+
})
|
32 |
+
|
33 |
+
# load pretrained model or not
|
34 |
+
trained_model_path = ''
|
35 |
+
load_state_dict(trained_model_path, model)
|
36 |
+
|
37 |
+
test_criterion = losses.__dict__['CELoss']()
|
38 |
+
|
39 |
+
test_dataset = ILSVRC2012Dataset(
|
40 |
+
root_dir=ILSVRC2012_path,
|
41 |
+
set_name='val',
|
42 |
+
transform=transforms.Compose([
|
43 |
+
Opencv2PIL(),
|
44 |
+
TorchResize(resize=input_image_size * scale),
|
45 |
+
TorchCenterCrop(resize=input_image_size),
|
46 |
+
TorchMeanStdNormalize(mean=[0.485, 0.456, 0.406],
|
47 |
+
std=[0.229, 0.224, 0.225]),
|
48 |
+
]))
|
49 |
+
test_collater = ClassificationCollater()
|
50 |
+
|
51 |
+
seed = 0
|
52 |
+
# batch_size is total size
|
53 |
+
batch_size = 256
|
54 |
+
# num_workers is total workers
|
55 |
+
num_workers = 16
|
imagenet/van_b0/train.sh
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.run --nproc_per_node=2 --master_addr 127.0.1.0 --master_port 10000 ../../../tools/train_classification_model.py --work-dir ./
|
imagenet/van_b0/train_config.py
ADDED
@@ -0,0 +1,126 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import sys
|
3 |
+
|
4 |
+
BASE_DIR = os.path.dirname(
|
5 |
+
os.path.dirname(os.path.dirname(os.path.dirname(
|
6 |
+
os.path.abspath(__file__)))))
|
7 |
+
sys.path.append(BASE_DIR)
|
8 |
+
|
9 |
+
from tools.path import ILSVRC2012_path
|
10 |
+
|
11 |
+
from simpleAICV.classification import backbones
|
12 |
+
from simpleAICV.classification import losses
|
13 |
+
from simpleAICV.classification.datasets.ilsvrc2012dataset import ILSVRC2012Dataset
|
14 |
+
from simpleAICV.classification.common import Opencv2PIL, TorchRandomResizedCrop, TorchRandomHorizontalFlip, RandAugment, TorchResize, TorchCenterCrop, TorchMeanStdNormalize, RandomErasing, ClassificationCollater, MixupCutmixClassificationCollater, load_state_dict
|
15 |
+
|
16 |
+
import torch
|
17 |
+
import torchvision.transforms as transforms
|
18 |
+
|
19 |
+
|
20 |
+
class config:
|
21 |
+
'''
|
22 |
+
for resnet,input_image_size = 224;for darknet,input_image_size = 256
|
23 |
+
'''
|
24 |
+
network = 'van_b0'
|
25 |
+
num_classes = 1000
|
26 |
+
input_image_size = 224
|
27 |
+
scale = 256 / 224
|
28 |
+
|
29 |
+
model = backbones.__dict__[network](**{
|
30 |
+
'drop_path_prob': 0.1,
|
31 |
+
'num_classes': num_classes,
|
32 |
+
})
|
33 |
+
|
34 |
+
# load pretrained model or not
|
35 |
+
trained_model_path = '/root/code/SimpleAICV_pytorch_training_examples_on_ImageNet_COCO_ADE20K/pretrained_models/van_weight_convert_from_official_weights/van_b0_pytorch_official_weight_convert.pth'
|
36 |
+
load_state_dict(trained_model_path, model)
|
37 |
+
|
38 |
+
train_criterion = losses.__dict__['OneHotLabelCELoss']()
|
39 |
+
test_criterion = losses.__dict__['CELoss']()
|
40 |
+
|
41 |
+
train_dataset = ILSVRC2012Dataset(
|
42 |
+
root_dir=ILSVRC2012_path,
|
43 |
+
set_name='train',
|
44 |
+
transform=transforms.Compose([
|
45 |
+
Opencv2PIL(),
|
46 |
+
TorchRandomResizedCrop(resize=input_image_size),
|
47 |
+
TorchRandomHorizontalFlip(prob=0.5),
|
48 |
+
RandAugment(magnitude=9,
|
49 |
+
num_layers=2,
|
50 |
+
resize=input_image_size,
|
51 |
+
mean=[0.485, 0.456, 0.406],
|
52 |
+
integer=True,
|
53 |
+
weight_idx=None,
|
54 |
+
magnitude_std=0.5,
|
55 |
+
magnitude_max=None),
|
56 |
+
TorchMeanStdNormalize(mean=[0.485, 0.456, 0.406],
|
57 |
+
std=[0.229, 0.224, 0.225]),
|
58 |
+
RandomErasing(prob=0.25, mode='pixel', max_count=1),
|
59 |
+
]))
|
60 |
+
|
61 |
+
test_dataset = ILSVRC2012Dataset(
|
62 |
+
root_dir=ILSVRC2012_path,
|
63 |
+
set_name='val',
|
64 |
+
transform=transforms.Compose([
|
65 |
+
Opencv2PIL(),
|
66 |
+
TorchResize(resize=input_image_size * scale),
|
67 |
+
TorchCenterCrop(resize=input_image_size),
|
68 |
+
TorchMeanStdNormalize(mean=[0.485, 0.456, 0.406],
|
69 |
+
std=[0.229, 0.224, 0.225]),
|
70 |
+
]))
|
71 |
+
|
72 |
+
train_collater = MixupCutmixClassificationCollater(
|
73 |
+
use_mixup=True,
|
74 |
+
mixup_alpha=0.8,
|
75 |
+
cutmix_alpha=1.0,
|
76 |
+
cutmix_minmax=None,
|
77 |
+
mixup_cutmix_prob=1.0,
|
78 |
+
switch_to_cutmix_prob=0.5,
|
79 |
+
mode='batch',
|
80 |
+
correct_lam=True,
|
81 |
+
label_smoothing=0.1,
|
82 |
+
num_classes=1000)
|
83 |
+
test_collater = ClassificationCollater()
|
84 |
+
|
85 |
+
seed = 0
|
86 |
+
# batch_size is total size
|
87 |
+
batch_size = 256
|
88 |
+
# num_workers is total workers
|
89 |
+
num_workers = 30
|
90 |
+
accumulation_steps = 4
|
91 |
+
|
92 |
+
optimizer = (
|
93 |
+
'AdamW',
|
94 |
+
{
|
95 |
+
'lr': 1e-4,
|
96 |
+
'global_weight_decay': False,
|
97 |
+
# if global_weight_decay = False
|
98 |
+
# all bias, bn and other 1d params weight set to 0 weight decay
|
99 |
+
'weight_decay': 1e-4,
|
100 |
+
'no_weight_decay_layer_name_list': [],
|
101 |
+
},
|
102 |
+
)
|
103 |
+
|
104 |
+
scheduler = (
|
105 |
+
'CosineLR',
|
106 |
+
{
|
107 |
+
'warm_up_epochs': 5,
|
108 |
+
'min_lr': 1e-6,
|
109 |
+
},
|
110 |
+
)
|
111 |
+
|
112 |
+
epochs = 300
|
113 |
+
print_interval = 50
|
114 |
+
|
115 |
+
sync_bn = False
|
116 |
+
use_amp = False
|
117 |
+
use_compile = False
|
118 |
+
compile_params = {
|
119 |
+
# 'default': optimizes for large models, low compile-time and no extra memory usage.
|
120 |
+
# 'reduce-overhead': optimizes to reduce the framework overhead and uses some extra memory, helps speed up small models, model update may not correct.
|
121 |
+
# 'max-autotune': optimizes to produce the fastest model, but takes a very long time to compile and may failed.
|
122 |
+
'mode': 'default',
|
123 |
+
}
|
124 |
+
|
125 |
+
use_ema_model = False
|
126 |
+
ema_model_decay = 0.9999
|
imagenet/van_b1/__pycache__/train_config.cpython-38.pyc
ADDED
Binary file (3.15 kB). View file
|
|
imagenet/van_b1/checkpoints/latest.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:39b1c45b4bc52676f3e15de8dafb4efb7ccb837e937024ead4d2aecfc7558a54
|
3 |
+
size 166723007
|
imagenet/van_b1/checkpoints/van_b1-acc80.956.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a7d54257e90cba48e94e2abcc8b39a24d26096eac4d18122e758ba4d0bc4e3f3
|
3 |
+
size 55600905
|
imagenet/van_b1/log/train.info.log
ADDED
The diff for this file is too large to render.
See raw diff
|
|
imagenet/van_b1/log/train.info.log.2023-11-28
ADDED
The diff for this file is too large to render.
See raw diff
|
|
imagenet/van_b1/test.sh
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.run --nproc_per_node=2 --master_addr 127.0.1.0 --master_port 10000 ../../../tools/test_classification_model.py --work-dir ./
|
imagenet/van_b1/test_config.py
ADDED
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import sys
|
3 |
+
|
4 |
+
BASE_DIR = os.path.dirname(
|
5 |
+
os.path.dirname(os.path.dirname(os.path.dirname(
|
6 |
+
os.path.abspath(__file__)))))
|
7 |
+
sys.path.append(BASE_DIR)
|
8 |
+
|
9 |
+
from tools.path import ILSVRC2012_path
|
10 |
+
|
11 |
+
from simpleAICV.classification import backbones
|
12 |
+
from simpleAICV.classification import losses
|
13 |
+
from simpleAICV.classification.datasets.ilsvrc2012dataset import ILSVRC2012Dataset
|
14 |
+
from simpleAICV.classification.common import Opencv2PIL, TorchResize, TorchCenterCrop, TorchMeanStdNormalize, ClassificationCollater, load_state_dict
|
15 |
+
|
16 |
+
import torch
|
17 |
+
import torchvision.transforms as transforms
|
18 |
+
|
19 |
+
|
20 |
+
class config:
|
21 |
+
'''
|
22 |
+
for resnet,input_image_size = 224;for darknet,input_image_size = 256
|
23 |
+
'''
|
24 |
+
network = 'van_b1'
|
25 |
+
num_classes = 1000
|
26 |
+
input_image_size = 224
|
27 |
+
scale = 256 / 224
|
28 |
+
|
29 |
+
model = backbones.__dict__[network](**{
|
30 |
+
'num_classes': num_classes,
|
31 |
+
})
|
32 |
+
|
33 |
+
# load pretrained model or not
|
34 |
+
trained_model_path = ''
|
35 |
+
load_state_dict(trained_model_path, model)
|
36 |
+
|
37 |
+
test_criterion = losses.__dict__['CELoss']()
|
38 |
+
|
39 |
+
test_dataset = ILSVRC2012Dataset(
|
40 |
+
root_dir=ILSVRC2012_path,
|
41 |
+
set_name='val',
|
42 |
+
transform=transforms.Compose([
|
43 |
+
Opencv2PIL(),
|
44 |
+
TorchResize(resize=input_image_size * scale),
|
45 |
+
TorchCenterCrop(resize=input_image_size),
|
46 |
+
TorchMeanStdNormalize(mean=[0.485, 0.456, 0.406],
|
47 |
+
std=[0.229, 0.224, 0.225]),
|
48 |
+
]))
|
49 |
+
test_collater = ClassificationCollater()
|
50 |
+
|
51 |
+
seed = 0
|
52 |
+
# batch_size is total size
|
53 |
+
batch_size = 256
|
54 |
+
# num_workers is total workers
|
55 |
+
num_workers = 16
|
imagenet/van_b1/train.sh
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.run --nproc_per_node=2 --master_addr 127.0.1.0 --master_port 10000 ../../../tools/train_classification_model.py --work-dir ./
|
imagenet/van_b1/train_config.py
ADDED
@@ -0,0 +1,126 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import sys
|
3 |
+
|
4 |
+
BASE_DIR = os.path.dirname(
|
5 |
+
os.path.dirname(os.path.dirname(os.path.dirname(
|
6 |
+
os.path.abspath(__file__)))))
|
7 |
+
sys.path.append(BASE_DIR)
|
8 |
+
|
9 |
+
from tools.path import ILSVRC2012_path
|
10 |
+
|
11 |
+
from simpleAICV.classification import backbones
|
12 |
+
from simpleAICV.classification import losses
|
13 |
+
from simpleAICV.classification.datasets.ilsvrc2012dataset import ILSVRC2012Dataset
|
14 |
+
from simpleAICV.classification.common import Opencv2PIL, TorchRandomResizedCrop, TorchRandomHorizontalFlip, RandAugment, TorchResize, TorchCenterCrop, TorchMeanStdNormalize, RandomErasing, ClassificationCollater, MixupCutmixClassificationCollater, load_state_dict
|
15 |
+
|
16 |
+
import torch
|
17 |
+
import torchvision.transforms as transforms
|
18 |
+
|
19 |
+
|
20 |
+
class config:
|
21 |
+
'''
|
22 |
+
for resnet,input_image_size = 224;for darknet,input_image_size = 256
|
23 |
+
'''
|
24 |
+
network = 'van_b1'
|
25 |
+
num_classes = 1000
|
26 |
+
input_image_size = 224
|
27 |
+
scale = 256 / 224
|
28 |
+
|
29 |
+
model = backbones.__dict__[network](**{
|
30 |
+
'drop_path_prob': 0.1,
|
31 |
+
'num_classes': num_classes,
|
32 |
+
})
|
33 |
+
|
34 |
+
# load pretrained model or not
|
35 |
+
trained_model_path = '/root/code/SimpleAICV_pytorch_training_examples_on_ImageNet_COCO_ADE20K/pretrained_models/van_weight_convert_from_official_weights/van_b1_pytorch_official_weight_convert.pth'
|
36 |
+
load_state_dict(trained_model_path, model)
|
37 |
+
|
38 |
+
train_criterion = losses.__dict__['OneHotLabelCELoss']()
|
39 |
+
test_criterion = losses.__dict__['CELoss']()
|
40 |
+
|
41 |
+
train_dataset = ILSVRC2012Dataset(
|
42 |
+
root_dir=ILSVRC2012_path,
|
43 |
+
set_name='train',
|
44 |
+
transform=transforms.Compose([
|
45 |
+
Opencv2PIL(),
|
46 |
+
TorchRandomResizedCrop(resize=input_image_size),
|
47 |
+
TorchRandomHorizontalFlip(prob=0.5),
|
48 |
+
RandAugment(magnitude=9,
|
49 |
+
num_layers=2,
|
50 |
+
resize=input_image_size,
|
51 |
+
mean=[0.485, 0.456, 0.406],
|
52 |
+
integer=True,
|
53 |
+
weight_idx=None,
|
54 |
+
magnitude_std=0.5,
|
55 |
+
magnitude_max=None),
|
56 |
+
TorchMeanStdNormalize(mean=[0.485, 0.456, 0.406],
|
57 |
+
std=[0.229, 0.224, 0.225]),
|
58 |
+
RandomErasing(prob=0.25, mode='pixel', max_count=1),
|
59 |
+
]))
|
60 |
+
|
61 |
+
test_dataset = ILSVRC2012Dataset(
|
62 |
+
root_dir=ILSVRC2012_path,
|
63 |
+
set_name='val',
|
64 |
+
transform=transforms.Compose([
|
65 |
+
Opencv2PIL(),
|
66 |
+
TorchResize(resize=input_image_size * scale),
|
67 |
+
TorchCenterCrop(resize=input_image_size),
|
68 |
+
TorchMeanStdNormalize(mean=[0.485, 0.456, 0.406],
|
69 |
+
std=[0.229, 0.224, 0.225]),
|
70 |
+
]))
|
71 |
+
|
72 |
+
train_collater = MixupCutmixClassificationCollater(
|
73 |
+
use_mixup=True,
|
74 |
+
mixup_alpha=0.8,
|
75 |
+
cutmix_alpha=1.0,
|
76 |
+
cutmix_minmax=None,
|
77 |
+
mixup_cutmix_prob=1.0,
|
78 |
+
switch_to_cutmix_prob=0.5,
|
79 |
+
mode='batch',
|
80 |
+
correct_lam=True,
|
81 |
+
label_smoothing=0.1,
|
82 |
+
num_classes=1000)
|
83 |
+
test_collater = ClassificationCollater()
|
84 |
+
|
85 |
+
seed = 0
|
86 |
+
# batch_size is total size
|
87 |
+
batch_size = 256
|
88 |
+
# num_workers is total workers
|
89 |
+
num_workers = 30
|
90 |
+
accumulation_steps = 4
|
91 |
+
|
92 |
+
optimizer = (
|
93 |
+
'AdamW',
|
94 |
+
{
|
95 |
+
'lr': 1e-4,
|
96 |
+
'global_weight_decay': False,
|
97 |
+
# if global_weight_decay = False
|
98 |
+
# all bias, bn and other 1d params weight set to 0 weight decay
|
99 |
+
'weight_decay': 1e-4,
|
100 |
+
'no_weight_decay_layer_name_list': [],
|
101 |
+
},
|
102 |
+
)
|
103 |
+
|
104 |
+
scheduler = (
|
105 |
+
'CosineLR',
|
106 |
+
{
|
107 |
+
'warm_up_epochs': 5,
|
108 |
+
'min_lr': 1e-6,
|
109 |
+
},
|
110 |
+
)
|
111 |
+
|
112 |
+
epochs = 300
|
113 |
+
print_interval = 50
|
114 |
+
|
115 |
+
sync_bn = False
|
116 |
+
use_amp = False
|
117 |
+
use_compile = False
|
118 |
+
compile_params = {
|
119 |
+
# 'default': optimizes for large models, low compile-time and no extra memory usage.
|
120 |
+
# 'reduce-overhead': optimizes to reduce the framework overhead and uses some extra memory, helps speed up small models, model update may not correct.
|
121 |
+
# 'max-autotune': optimizes to produce the fastest model, but takes a very long time to compile and may failed.
|
122 |
+
'mode': 'default',
|
123 |
+
}
|
124 |
+
|
125 |
+
use_ema_model = False
|
126 |
+
ema_model_decay = 0.9999
|
imagenet/van_b2/__pycache__/train_config.cpython-38.pyc
ADDED
Binary file (3.15 kB). View file
|
|
imagenet/van_b2/checkpoints/latest.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f7bb5fa0d4292847ea6e686273c143b2d8efb4b9a79ff97c63dbfc837579c38f
|
3 |
+
size 319694059
|
imagenet/van_b2/checkpoints/van_b2-acc82.322.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7d62b46a66177819c875a50822bbda4190cac5c989cecde2cd1963b68e6c6f7d
|
3 |
+
size 106609413
|
imagenet/van_b2/log/train.info.log
ADDED
The diff for this file is too large to render.
See raw diff
|
|
imagenet/van_b2/log/train.info.log.2023-11-21
ADDED
The diff for this file is too large to render.
See raw diff
|
|
imagenet/van_b2/test.sh
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.run --nproc_per_node=2 --master_addr 127.0.1.0 --master_port 10000 ../../../tools/test_classification_model.py --work-dir ./
|
imagenet/van_b2/test_config.py
ADDED
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import sys
|
3 |
+
|
4 |
+
BASE_DIR = os.path.dirname(
|
5 |
+
os.path.dirname(os.path.dirname(os.path.dirname(
|
6 |
+
os.path.abspath(__file__)))))
|
7 |
+
sys.path.append(BASE_DIR)
|
8 |
+
|
9 |
+
from tools.path import ILSVRC2012_path
|
10 |
+
|
11 |
+
from simpleAICV.classification import backbones
|
12 |
+
from simpleAICV.classification import losses
|
13 |
+
from simpleAICV.classification.datasets.ilsvrc2012dataset import ILSVRC2012Dataset
|
14 |
+
from simpleAICV.classification.common import Opencv2PIL, TorchResize, TorchCenterCrop, TorchMeanStdNormalize, ClassificationCollater, load_state_dict
|
15 |
+
|
16 |
+
import torch
|
17 |
+
import torchvision.transforms as transforms
|
18 |
+
|
19 |
+
|
20 |
+
class config:
|
21 |
+
'''
|
22 |
+
for resnet,input_image_size = 224;for darknet,input_image_size = 256
|
23 |
+
'''
|
24 |
+
network = 'van_b2'
|
25 |
+
num_classes = 1000
|
26 |
+
input_image_size = 224
|
27 |
+
scale = 256 / 224
|
28 |
+
|
29 |
+
model = backbones.__dict__[network](**{
|
30 |
+
'num_classes': num_classes,
|
31 |
+
})
|
32 |
+
|
33 |
+
# load pretrained model or not
|
34 |
+
trained_model_path = ''
|
35 |
+
load_state_dict(trained_model_path, model)
|
36 |
+
|
37 |
+
test_criterion = losses.__dict__['CELoss']()
|
38 |
+
|
39 |
+
test_dataset = ILSVRC2012Dataset(
|
40 |
+
root_dir=ILSVRC2012_path,
|
41 |
+
set_name='val',
|
42 |
+
transform=transforms.Compose([
|
43 |
+
Opencv2PIL(),
|
44 |
+
TorchResize(resize=input_image_size * scale),
|
45 |
+
TorchCenterCrop(resize=input_image_size),
|
46 |
+
TorchMeanStdNormalize(mean=[0.485, 0.456, 0.406],
|
47 |
+
std=[0.229, 0.224, 0.225]),
|
48 |
+
]))
|
49 |
+
test_collater = ClassificationCollater()
|
50 |
+
|
51 |
+
seed = 0
|
52 |
+
# batch_size is total size
|
53 |
+
batch_size = 256
|
54 |
+
# num_workers is total workers
|
55 |
+
num_workers = 16
|
imagenet/van_b2/train.sh
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
CUDA_VISIBLE_DEVICES=0,1,2,3 python -m torch.distributed.run --nproc_per_node=4 --master_addr 127.0.1.0 --master_port 10000 ../../../tools/train_classification_model.py --work-dir ./
|
imagenet/van_b2/train_config.py
ADDED
@@ -0,0 +1,126 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import sys
|
3 |
+
|
4 |
+
BASE_DIR = os.path.dirname(
|
5 |
+
os.path.dirname(os.path.dirname(os.path.dirname(
|
6 |
+
os.path.abspath(__file__)))))
|
7 |
+
sys.path.append(BASE_DIR)
|
8 |
+
|
9 |
+
from tools.path import ILSVRC2012_path
|
10 |
+
|
11 |
+
from simpleAICV.classification import backbones
|
12 |
+
from simpleAICV.classification import losses
|
13 |
+
from simpleAICV.classification.datasets.ilsvrc2012dataset import ILSVRC2012Dataset
|
14 |
+
from simpleAICV.classification.common import Opencv2PIL, TorchRandomResizedCrop, TorchRandomHorizontalFlip, RandAugment, TorchResize, TorchCenterCrop, TorchMeanStdNormalize, RandomErasing, ClassificationCollater, MixupCutmixClassificationCollater, load_state_dict
|
15 |
+
|
16 |
+
import torch
|
17 |
+
import torchvision.transforms as transforms
|
18 |
+
|
19 |
+
|
20 |
+
class config:
|
21 |
+
'''
|
22 |
+
for resnet,input_image_size = 224;for darknet,input_image_size = 256
|
23 |
+
'''
|
24 |
+
network = 'van_b2'
|
25 |
+
num_classes = 1000
|
26 |
+
input_image_size = 224
|
27 |
+
scale = 256 / 224
|
28 |
+
|
29 |
+
model = backbones.__dict__[network](**{
|
30 |
+
'drop_path_prob': 0.1,
|
31 |
+
'num_classes': num_classes,
|
32 |
+
})
|
33 |
+
|
34 |
+
# load pretrained model or not
|
35 |
+
trained_model_path = '/root/code/SimpleAICV_pytorch_training_examples_on_ImageNet_COCO_ADE20K/pretrained_models/van_weight_convert_from_official_weights/van_b2_pytorch_official_weight_convert.pth'
|
36 |
+
load_state_dict(trained_model_path, model)
|
37 |
+
|
38 |
+
train_criterion = losses.__dict__['OneHotLabelCELoss']()
|
39 |
+
test_criterion = losses.__dict__['CELoss']()
|
40 |
+
|
41 |
+
train_dataset = ILSVRC2012Dataset(
|
42 |
+
root_dir=ILSVRC2012_path,
|
43 |
+
set_name='train',
|
44 |
+
transform=transforms.Compose([
|
45 |
+
Opencv2PIL(),
|
46 |
+
TorchRandomResizedCrop(resize=input_image_size),
|
47 |
+
TorchRandomHorizontalFlip(prob=0.5),
|
48 |
+
RandAugment(magnitude=9,
|
49 |
+
num_layers=2,
|
50 |
+
resize=input_image_size,
|
51 |
+
mean=[0.485, 0.456, 0.406],
|
52 |
+
integer=True,
|
53 |
+
weight_idx=None,
|
54 |
+
magnitude_std=0.5,
|
55 |
+
magnitude_max=None),
|
56 |
+
TorchMeanStdNormalize(mean=[0.485, 0.456, 0.406],
|
57 |
+
std=[0.229, 0.224, 0.225]),
|
58 |
+
RandomErasing(prob=0.25, mode='pixel', max_count=1),
|
59 |
+
]))
|
60 |
+
|
61 |
+
test_dataset = ILSVRC2012Dataset(
|
62 |
+
root_dir=ILSVRC2012_path,
|
63 |
+
set_name='val',
|
64 |
+
transform=transforms.Compose([
|
65 |
+
Opencv2PIL(),
|
66 |
+
TorchResize(resize=input_image_size * scale),
|
67 |
+
TorchCenterCrop(resize=input_image_size),
|
68 |
+
TorchMeanStdNormalize(mean=[0.485, 0.456, 0.406],
|
69 |
+
std=[0.229, 0.224, 0.225]),
|
70 |
+
]))
|
71 |
+
|
72 |
+
train_collater = MixupCutmixClassificationCollater(
|
73 |
+
use_mixup=True,
|
74 |
+
mixup_alpha=0.8,
|
75 |
+
cutmix_alpha=1.0,
|
76 |
+
cutmix_minmax=None,
|
77 |
+
mixup_cutmix_prob=1.0,
|
78 |
+
switch_to_cutmix_prob=0.5,
|
79 |
+
mode='batch',
|
80 |
+
correct_lam=True,
|
81 |
+
label_smoothing=0.1,
|
82 |
+
num_classes=1000)
|
83 |
+
test_collater = ClassificationCollater()
|
84 |
+
|
85 |
+
seed = 0
|
86 |
+
# batch_size is total size
|
87 |
+
batch_size = 256
|
88 |
+
# num_workers is total workers
|
89 |
+
num_workers = 60
|
90 |
+
accumulation_steps = 4
|
91 |
+
|
92 |
+
optimizer = (
|
93 |
+
'AdamW',
|
94 |
+
{
|
95 |
+
'lr': 1e-4,
|
96 |
+
'global_weight_decay': False,
|
97 |
+
# if global_weight_decay = False
|
98 |
+
# all bias, bn and other 1d params weight set to 0 weight decay
|
99 |
+
'weight_decay': 1e-4,
|
100 |
+
'no_weight_decay_layer_name_list': [],
|
101 |
+
},
|
102 |
+
)
|
103 |
+
|
104 |
+
scheduler = (
|
105 |
+
'CosineLR',
|
106 |
+
{
|
107 |
+
'warm_up_epochs': 5,
|
108 |
+
'min_lr': 1e-6,
|
109 |
+
},
|
110 |
+
)
|
111 |
+
|
112 |
+
epochs = 300
|
113 |
+
print_interval = 50
|
114 |
+
|
115 |
+
sync_bn = False
|
116 |
+
use_amp = False
|
117 |
+
use_compile = False
|
118 |
+
compile_params = {
|
119 |
+
# 'default': optimizes for large models, low compile-time and no extra memory usage.
|
120 |
+
# 'reduce-overhead': optimizes to reduce the framework overhead and uses some extra memory, helps speed up small models, model update may not correct.
|
121 |
+
# 'max-autotune': optimizes to produce the fastest model, but takes a very long time to compile and may failed.
|
122 |
+
'mode': 'default',
|
123 |
+
}
|
124 |
+
|
125 |
+
use_ema_model = False
|
126 |
+
ema_model_decay = 0.9999
|