训练报错

[zbl929]

norm_cfg = dict(type='BN', requires_grad=True)
model = dict(
type='EncoderDecoder',
backbone=dict(
delete=True,
type='Vanillanet',
act_num=3,
dims=[512, 512, 1024, 2048, 2048, 2048, 2048, 2048, 2048, 4096, 4096],
strides=[1, 2, 2, 1, 1, 1, 1, 1, 2, 1],
out_indices=[1, 2, 8, 9],
norm_cfg=dict(type='BN', requires_grad=True),
with_extra_norm=True),
neck=dict(
type='FPN',
in_channels=[512, 1024, 2048, 4096],
out_channels=1024,
num_outs=4),
decode_head=dict(
type='SegformerHead',
in_channels=[128, 256, 512, 1024],
in_index=[0, 1, 2, 3],
channels=512,
dropout_ratio=0.1,
num_classes=2,
norm_cfg=dict(type='BN', requires_grad=True),
align_corners=False,
loss_decode=dict(
type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0)),
train_cfg=dict(),
test_cfg=dict(mode='whole'))
dataset_type = 'WaterDataset'
data_root = 'E:\zbl\pic\segformer-pytorch-master\VOCdevkit\zbl'
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
crop_size = (512, 512)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='LoadAnnotations', reduce_zero_label=False),
dict(type='Resize', img_scale=(512, 512), ratio_range=(0.5, 2.0)),
dict(type='RandomCrop', crop_size=(512, 512), cat_max_ratio=0.75),
dict(type='RandomFlip', prob=0.5),
dict(type='PhotoMetricDistortion'),
dict(
type='Normalize',
mean=[123.675, 116.28, 103.53],
std=[58.395, 57.12, 57.375],
to_rgb=True),
dict(type='Pad', size=(512, 512), pad_val=0, seg_pad_val=255),
dict(type='DefaultFormatBundle'),
dict(type='Collect', keys=['img', 'gt_semantic_seg'])
]
test_pipeline = [
dict(type='LoadImageFromFile'),
dict(
type='MultiScaleFlipAug',
img_scale=(512, 512),
flip=False,
transforms=[
dict(type='Resize', keep_ratio=True),
dict(type='RandomFlip'),
dict(
type='Normalize',
mean=[123.675, 116.28, 103.53],
std=[58.395, 57.12, 57.375],
to_rgb=True),
dict(type='ImageToTensor', keys=['img']),
dict(type='Collect', keys=['img'])
])
]
data = dict(
samples_per_gpu=8,
workers_per_gpu=2,
train=dict(
type='WaterDataset',
data_root='E:\zbl\pic\segformer-pytorch-master\VOCdevkit\zbl',
img_dir='JPEGImages',
ann_dir='SegmentationClass',
split='ImageSets/Segmentation/train.txt',
pipeline=[
dict(type='LoadImageFromFile'),
dict(type='LoadAnnotations', reduce_zero_label=False),
dict(type='Resize', img_scale=(512, 512), ratio_range=(0.5, 2.0)),
dict(type='RandomCrop', crop_size=(512, 512), cat_max_ratio=0.75),
dict(type='RandomFlip', prob=0.5),
dict(type='PhotoMetricDistortion'),
dict(
type='Normalize',
mean=[123.675, 116.28, 103.53],
std=[58.395, 57.12, 57.375],
to_rgb=True),
dict(type='Pad', size=(512, 512), pad_val=0, seg_pad_val=255),
dict(type='DefaultFormatBundle'),
dict(type='Collect', keys=['img', 'gt_semantic_seg'])
]),
val=dict(
type='WaterDataset',
data_root='E:\zbl\pic\segformer-pytorch-master\VOCdevkit\zbl',
img_dir='JPEGImages',
ann_dir='SegmentationClass',
split='ImageSets/Segmentation/val.txt',
pipeline=[
dict(type='LoadImageFromFile'),
dict(
type='MultiScaleFlipAug',
img_scale=(512, 512),
flip=False,
transforms=[
dict(type='Resize', keep_ratio=True),
dict(type='RandomFlip'),
dict(
type='Normalize',
mean=[123.675, 116.28, 103.53],
std=[58.395, 57.12, 57.375],
to_rgb=True),
dict(type='ImageToTensor', keys=['img']),
dict(type='Collect', keys=['img'])
])
]),
test=dict(
type='WaterDataset',
data_root='E:\zbl\pic\segformer-pytorch-master\VOCdevkit\zbl',
img_dir='JPEGImages',
ann_dir='SegmentationClass',
split='ImageSets/Segmentation/val.txt',
pipeline=[
dict(type='LoadImageFromFile'),
dict(
type='MultiScaleFlipAug',
img_scale=(512, 512),
flip=False,
transforms=[
dict(type='Resize', keep_ratio=True),
dict(type='RandomFlip'),
dict(
type='Normalize',
mean=[123.675, 116.28, 103.53],
std=[58.395, 57.12, 57.375],
to_rgb=True),
dict(type='ImageToTensor', keys=['img']),
dict(type='Collect', keys=['img'])
])
]))
log_config = dict(
interval=50, hooks=[dict(type='TextLoggerHook', by_epoch=False)])
dist_params = dict(backend='nccl')
log_level = 'INFO'
load_from = 'E:\zbl\mmsegmentation-0.23.0\pretrain\vanillanet_12.pth'
resume_from = None
workflow = [('train', 1), ('val', 1)]
cudnn_benchmark = True
optimizer = dict(
constructor='LearningRateDecayOptimizerConstructor',
type='AdamW',
lr=0.00013,
betas=(0.9, 0.999),
weight_decay=0.05,
paramwise_cfg=dict(decay_rate=0.6, decay_type='layer_wise', num_layers=6))
optimizer_config = dict(grad_clip=None)
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[15000, 20000])
runner = dict(type='IterBasedRunner', max_iters=20000)
checkpoint_config = dict(by_epoch=False, interval=2000)
evaluation = dict(interval=2000, metric='mIoU')
work_dir = 'logs_vanillanet/pretrain_12'
gpu_ids = [0]
auto_resume = False

---

报错信息
Traceback (most recent call last):
File "E:/zbl/mmsegmentation-0.23.0/tools/train.py", line 244, in
main()
File "E:/zbl/mmsegmentation-0.23.0/tools/train.py", line 233, in main
train_segmentor(
File "E:\zbl\mmsegmentation-0.23.0\mmseg\apis\train.py", line 174, in train_segmentor
runner.run(data_loaders, cfg.workflow)
File "D:\Anaconda3.0\envs\nextvit\lib\site-packages\mmcv\runner\iter_based_runner.py", line 134, in run
iter_runner(iter_loaders[i], **kwargs)
File "D:\Anaconda3.0\envs\nextvit\lib\site-packages\mmcv\runner\iter_based_runner.py", line 61, in train
outputs = self.model.train_step(data_batch, self.optimizer, **kwargs)
File "D:\Anaconda3.0\envs\nextvit\lib\site-packages\mmcv\parallel\data_parallel.py", line 75, in train_step
return self.module.train_step(*inputs[0], **kwargs[0])
File "E:\zbl\mmsegmentation-0.23.0\mmseg\models\segmentors\base.py", line 138, in train_step
losses = self(**data_batch)
File "D:\Anaconda3.0\envs\nextvit\lib\site-packages\torch\nn\modules\module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "D:\Anaconda3.0\envs\nextvit\lib\site-packages\mmcv\runner\fp16_utils.py", line 110, in new_func
return old_func(*args, **kwargs)
File "E:\zbl\mmsegmentation-0.23.0\mmseg\models\segmentors\base.py", line 108, in forward
return self.forward_train(img, img_metas, **kwargs)
File "E:\zbl\mmsegmentation-0.23.0\mmseg\models\segmentors\encoder_decoder.py", line 139, in forward_train
x = self.extract_feat(img)
File "E:\zbl\mmsegmentation-0.23.0\mmseg\models\segmentors\encoder_decoder.py", line 67, in extract_feat
x = self.neck(x)
File "D:\Anaconda3.0\envs\nextvit\lib\site-packages\torch\nn\modules\module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "D:\Anaconda3.0\envs\nextvit\lib\site-packages\mmcv\runner\fp16_utils.py", line 110, in new_func
return old_func(*args, **kwargs)
File "E:\zbl\mmsegmentation-0.23.0\mmseg\models\necks\fpn.py", line 167, in forward
laterals = [
File "E:\zbl\mmsegmentation-0.23.0\mmseg\models\necks\fpn.py", line 168, in
lateral_conv(inputs[i + self.start_level])
File "D:\Anaconda3.0\envs\nextvit\lib\site-packages\torch\nn\modules\module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "D:\Anaconda3.0\envs\nextvit\lib\site-packages\mmcv\cnn\bricks\conv_module.py", line 201, in forward
x = self.conv(x)
File "D:\Anaconda3.0\envs\nextvit\lib\site-packages\torch\nn\modules\module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "D:\Anaconda3.0\envs\nextvit\lib\site-packages\torch\nn\modules\conv.py", line 446, in forward
return self._conv_forward(input, self.weight, self.bias)
File "D:\Anaconda3.0\envs\nextvit\lib\site-packages\torch\nn\modules\conv.py", line 442, in _conv_forward
return F.conv2d(input, weight, bias, self.stride,
RuntimeError: Given groups=1, weight of size [1024, 512, 1, 1], expected input[8, 1024, 64, 64] to have 512 channels, but got 1024 channels instead

进程已结束,退出代码1

---

试验了vanillanet12和vanillanet13，都报同样的错

期望尽快解答！感谢！

[ggjy]

hi, 看起来你的out_indices=[1, 2, 8, 9]应该改成out_indices=[0, 1, 7, 9]，不然backbone其实return的4个stage的channel就是[1024, 2048, 4096, 4096]，和FPN的输入对不上。

[zbl929]

您好，感谢回答！！
我现在又碰上了另一个问题：加载预训练权重无法匹配！无论是从load_from加载或是从配置文件model里加载，都显示WARNING - No pre-trained weights for Vanillanet, training start from scratch！
期待解答！感谢！

---

2023-06-30 16:38:52,882 - mmseg - INFO - OpenCV num_threads is `
2023-06-30 16:38:58,558 - mmseg - INFO - Distributed training: False
2023-06-30 16:38:59,138 - mmseg - INFO - Config:
norm_cfg = dict(type='BN', requires_grad=True)
model = dict(
type='EncoderDecoder',
pretrained='E:/zbl/mmsegmentation-0.23.0/pretrain/vanillanet_13.pth',
backbone=dict(
delete=True,
type='Vanillanet',
act_num=4,
dims=[
512, 512, 1024, 2048, 2048, 2048, 2048, 2048, 2048, 2048, 4096,
4096
],
strides=[1, 2, 2, 1, 1, 1, 1, 1, 1, 2, 1],
out_indices=[0, 1, 8, 10],
norm_cfg=dict(type='BN', requires_grad=True),
with_extra_norm=True),
neck=dict(
type='FPN',
in_channels=[512, 1024, 2048, 4096],
out_channels=1024,
num_outs=4),
decode_head=dict(
type='FPNHead',
in_channels=[1024, 1024, 1024, 1024],
in_index=[0, 1, 2, 3],
feature_strides=[4, 8, 16, 32],
channels=512,
dropout_ratio=0.1,
num_classes=2,
norm_cfg=dict(type='BN', requires_grad=True),
align_corners=False,
loss_decode=dict(
type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
train_cfg=dict(),
test_cfg=dict(mode='whole'))
dataset_type = 'WaterDataset'
data_root = 'E:\zbl\pic\segformer-pytorch-master\VOCdevkit\zbl'
img_norm_cfg = dict(
mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
crop_size = (512, 512)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='LoadAnnotations', reduce_zero_label=False),
dict(type='Resize', img_scale=(512, 512), ratio_range=(0.5, 2.0)),
dict(type='RandomCrop', crop_size=(512, 512), cat_max_ratio=0.75),
dict(type='RandomFlip', prob=0.5),
dict(type='PhotoMetricDistortion'),
dict(
type='Normalize',
mean=[123.675, 116.28, 103.53],
std=[58.395, 57.12, 57.375],
to_rgb=True),
dict(type='Pad', size=(512, 512), pad_val=0, seg_pad_val=255),
dict(type='DefaultFormatBundle'),
dict(type='Collect', keys=['img', 'gt_semantic_seg'])
]
test_pipeline = [
dict(type='LoadImageFromFile'),
dict(
type='MultiScaleFlipAug',
img_scale=(512, 512),
flip=False,
transforms=[
dict(type='Resize', keep_ratio=True),
dict(type='RandomFlip'),
dict(
type='Normalize',
mean=[123.675, 116.28, 103.53],
std=[58.395, 57.12, 57.375],
to_rgb=True),
dict(type='ImageToTensor', keys=['img']),
dict(type='Collect', keys=['img'])
])
]
data = dict(
samples_per_gpu=8,
workers_per_gpu=1,
train=dict(
type='WaterDataset',
data_root='E:\zbl\pic\segformer-pytorch-master\VOCdevkit\zbl',
img_dir='JPEGImages',
ann_dir='SegmentationClass',
split='ImageSets/Segmentation/train.txt',
pipeline=[
dict(type='LoadImageFromFile'),
dict(type='LoadAnnotations', reduce_zero_label=False),
dict(type='Resize', img_scale=(512, 512), ratio_range=(0.5, 2.0)),
dict(type='RandomCrop', crop_size=(512, 512), cat_max_ratio=0.75),
dict(type='RandomFlip', prob=0.5),
dict(type='PhotoMetricDistortion'),
dict(
type='Normalize',
mean=[123.675, 116.28, 103.53],
std=[58.395, 57.12, 57.375],
to_rgb=True),
dict(type='Pad', size=(512, 512), pad_val=0, seg_pad_val=255),
dict(type='DefaultFormatBundle'),
dict(type='Collect', keys=['img', 'gt_semantic_seg'])
]),
val=dict(
type='WaterDataset',
data_root='E:\zbl\pic\segformer-pytorch-master\VOCdevkit\zbl',
img_dir='JPEGImages',
ann_dir='SegmentationClass',
split='ImageSets/Segmentation/val.txt',
pipeline=[
dict(type='LoadImageFromFile'),
dict(
type='MultiScaleFlipAug',
img_scale=(512, 512),
flip=False,
transforms=[
dict(type='Resize', keep_ratio=True),
dict(type='RandomFlip'),
dict(
type='Normalize',
mean=[123.675, 116.28, 103.53],
std=[58.395, 57.12, 57.375],
to_rgb=True),
dict(type='ImageToTensor', keys=['img']),
dict(type='Collect', keys=['img'])
])
]),
test=dict(
type='WaterDataset',
data_root='E:\zbl\pic\segformer-pytorch-master\VOCdevkit\zbl',
img_dir='JPEGImages',
ann_dir='SegmentationClass',
split='ImageSets/Segmentation/val.txt',
pipeline=[
dict(type='LoadImageFromFile'),
dict(
type='MultiScaleFlipAug',
img_scale=(512, 512),
flip=False,
transforms=[
dict(type='Resize', keep_ratio=True),
dict(type='RandomFlip'),
dict(
type='Normalize',
mean=[123.675, 116.28, 103.53],
std=[58.395, 57.12, 57.375],
to_rgb=True),
dict(type='ImageToTensor', keys=['img']),
dict(type='Collect', keys=['img'])
])
]))
log_config = dict(
interval=20, hooks=[dict(type='TextLoggerHook', by_epoch=False)])
dist_params = dict(backend='nccl')
log_level = 'INFO'
load_from = None
resume_from = None
workflow = [('train', 1), ('val', 1)]
cudnn_benchmark = True
optimizer = dict(
constructor='LearningRateDecayOptimizerConstructor',
type='AdamW',
lr=0.00013,
betas=(0.9, 0.999),
weight_decay=0.05,
paramwise_cfg=dict(decay_rate=0.6, decay_type='layer_wise', num_layers=6))
optimizer_config = dict(grad_clip=None)
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[15000, 20000])
runner = dict(type='IterBasedRunner', max_iters=20000)
checkpoint_config = dict(by_epoch=False, interval=2000)
evaluation = dict(interval=2000, metric='mIoU')
work_dir = 'E:\zbl\mmsegmentation-0.23.0\tools\logs_vanillanet\13'
gpu_ids = [0]
auto_resume = False

2023-06-30 16:38:59,145 - mmseg - INFO - Set random seed to 1502961241, deterministic: False
2023-06-30 16:39:01,437 - mmseg - WARNING - No pre-trained weights for Vanillanet, training start from scratch
2023-06-30 16:39:02,380 - mmseg - INFO - initialize FPN with init_cfg {'type': 'Xavier', 'layer': 'Conv2d', 'distribution': 'uniform'}
2023-06-30 16:39:02,785 - mmseg - INFO - initialize FPNHead with init_cfg {'type': 'Normal', 'std': 0.01, 'override': {'name': 'conv_seg'}}
Name of parameter - Initialization information

backbone.stem1.0.weight - torch.Size([512, 3, 4, 4]):
Initialized by user-defined init_weights in Vanillanet

backbone.stem1.0.bias - torch.Size([512]):
Initialized by user-defined init_weights in Vanillanet

backbone.stem1.1.weight - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stem1.1.bias - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stem2.0.weight - torch.Size([512, 512, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stem2.0.bias - torch.Size([512]):
Initialized by user-defined init_weights in Vanillanet

backbone.stem2.1.weight - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stem2.1.bias - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stem2.2.weight - torch.Size([512, 1, 9, 9]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stem2.2.bn.weight - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stem2.2.bn.bias - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.0.conv1.0.weight - torch.Size([512, 512, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.0.conv1.0.bias - torch.Size([512]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.0.conv1.1.weight - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.0.conv1.1.bias - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.0.conv2.0.weight - torch.Size([512, 512, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.0.conv2.0.bias - torch.Size([512]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.0.conv2.1.weight - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.0.conv2.1.bias - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.0.act.weight - torch.Size([512, 1, 9, 9]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.0.act.bn.weight - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.0.act.bn.bias - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.1.conv1.0.weight - torch.Size([512, 512, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.1.conv1.0.bias - torch.Size([512]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.1.conv1.1.weight - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.1.conv1.1.bias - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.1.conv2.0.weight - torch.Size([1024, 512, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.1.conv2.0.bias - torch.Size([1024]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.1.conv2.1.weight - torch.Size([1024]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.1.conv2.1.bias - torch.Size([1024]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.1.act.weight - torch.Size([1024, 1, 9, 9]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.1.act.bn.weight - torch.Size([1024]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.1.act.bn.bias - torch.Size([1024]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.2.conv1.0.weight - torch.Size([1024, 1024, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.2.conv1.0.bias - torch.Size([1024]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.2.conv1.1.weight - torch.Size([1024]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.2.conv1.1.bias - torch.Size([1024]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.2.conv2.0.weight - torch.Size([2048, 1024, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.2.conv2.0.bias - torch.Size([2048]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.2.conv2.1.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.2.conv2.1.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.2.act.weight - torch.Size([2048, 1, 9, 9]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.2.act.bn.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.2.act.bn.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.3.conv1.0.weight - torch.Size([2048, 2048, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.3.conv1.0.bias - torch.Size([2048]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.3.conv1.1.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.3.conv1.1.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.3.conv2.0.weight - torch.Size([2048, 2048, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.3.conv2.0.bias - torch.Size([2048]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.3.conv2.1.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.3.conv2.1.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.3.act.weight - torch.Size([2048, 1, 9, 9]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.3.act.bn.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.3.act.bn.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.4.conv1.0.weight - torch.Size([2048, 2048, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.4.conv1.0.bias - torch.Size([2048]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.4.conv1.1.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.4.conv1.1.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.4.conv2.0.weight - torch.Size([2048, 2048, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.4.conv2.0.bias - torch.Size([2048]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.4.conv2.1.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.4.conv2.1.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.4.act.weight - torch.Size([2048, 1, 9, 9]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.4.act.bn.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.4.act.bn.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.5.conv1.0.weight - torch.Size([2048, 2048, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.5.conv1.0.bias - torch.Size([2048]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.5.conv1.1.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.5.conv1.1.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.5.conv2.0.weight - torch.Size([2048, 2048, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.5.conv2.0.bias - torch.Size([2048]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.5.conv2.1.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.5.conv2.1.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.5.act.weight - torch.Size([2048, 1, 9, 9]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.5.act.bn.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.5.act.bn.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.6.conv1.0.weight - torch.Size([2048, 2048, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.6.conv1.0.bias - torch.Size([2048]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.6.conv1.1.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.6.conv1.1.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.6.conv2.0.weight - torch.Size([2048, 2048, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.6.conv2.0.bias - torch.Size([2048]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.6.conv2.1.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.6.conv2.1.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.6.act.weight - torch.Size([2048, 1, 9, 9]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.6.act.bn.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.6.act.bn.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.7.conv1.0.weight - torch.Size([2048, 2048, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.7.conv1.0.bias - torch.Size([2048]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.7.conv1.1.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.7.conv1.1.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.7.conv2.0.weight - torch.Size([2048, 2048, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.7.conv2.0.bias - torch.Size([2048]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.7.conv2.1.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.7.conv2.1.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.7.act.weight - torch.Size([2048, 1, 9, 9]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.7.act.bn.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.7.act.bn.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.8.conv1.0.weight - torch.Size([2048, 2048, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.8.conv1.0.bias - torch.Size([2048]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.8.conv1.1.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.8.conv1.1.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.8.conv2.0.weight - torch.Size([2048, 2048, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.8.conv2.0.bias - torch.Size([2048]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.8.conv2.1.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.8.conv2.1.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.8.act.weight - torch.Size([2048, 1, 9, 9]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.8.act.bn.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.8.act.bn.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.9.conv1.0.weight - torch.Size([2048, 2048, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.9.conv1.0.bias - torch.Size([2048]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.9.conv1.1.weight - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.9.conv1.1.bias - torch.Size([2048]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.9.conv2.0.weight - torch.Size([4096, 2048, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.9.conv2.0.bias - torch.Size([4096]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.9.conv2.1.weight - torch.Size([4096]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.9.conv2.1.bias - torch.Size([4096]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.9.act.weight - torch.Size([4096, 1, 9, 9]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.9.act.bn.weight - torch.Size([4096]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.9.act.bn.bias - torch.Size([4096]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.10.conv1.0.weight - torch.Size([4096, 4096, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.10.conv1.0.bias - torch.Size([4096]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.10.conv1.1.weight - torch.Size([4096]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.10.conv1.1.bias - torch.Size([4096]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.10.conv2.0.weight - torch.Size([4096, 4096, 1, 1]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.10.conv2.0.bias - torch.Size([4096]):
Initialized by user-defined init_weights in Vanillanet

backbone.stages.10.conv2.1.weight - torch.Size([4096]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.10.conv2.1.bias - torch.Size([4096]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.10.act.weight - torch.Size([4096, 1, 9, 9]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.10.act.bn.weight - torch.Size([4096]):
The value is the same before and after calling init_weights of EncoderDecoder

backbone.stages.10.act.bn.bias - torch.Size([4096]):
The value is the same before and after calling init_weights of EncoderDecoder

neck.lateral_convs.0.conv.weight - torch.Size([1024, 512, 1, 1]):
XavierInit: gain=1, distribution=uniform, bias=0

neck.lateral_convs.0.conv.bias - torch.Size([1024]):
The value is the same before and after calling init_weights of EncoderDecoder

neck.lateral_convs.1.conv.weight - torch.Size([1024, 1024, 1, 1]):
XavierInit: gain=1, distribution=uniform, bias=0

neck.lateral_convs.1.conv.bias - torch.Size([1024]):
The value is the same before and after calling init_weights of EncoderDecoder

neck.lateral_convs.2.conv.weight - torch.Size([1024, 2048, 1, 1]):
XavierInit: gain=1, distribution=uniform, bias=0

neck.lateral_convs.2.conv.bias - torch.Size([1024]):
The value is the same before and after calling init_weights of EncoderDecoder

neck.lateral_convs.3.conv.weight - torch.Size([1024, 4096, 1, 1]):
XavierInit: gain=1, distribution=uniform, bias=0

neck.lateral_convs.3.conv.bias - torch.Size([1024]):
The value is the same before and after calling init_weights of EncoderDecoder

neck.fpn_convs.0.conv.weight - torch.Size([1024, 1024, 3, 3]):
XavierInit: gain=1, distribution=uniform, bias=0

neck.fpn_convs.0.conv.bias - torch.Size([1024]):
The value is the same before and after calling init_weights of EncoderDecoder

neck.fpn_convs.1.conv.weight - torch.Size([1024, 1024, 3, 3]):
XavierInit: gain=1, distribution=uniform, bias=0

neck.fpn_convs.1.conv.bias - torch.Size([1024]):
The value is the same before and after calling init_weights of EncoderDecoder

neck.fpn_convs.2.conv.weight - torch.Size([1024, 1024, 3, 3]):
XavierInit: gain=1, distribution=uniform, bias=0

neck.fpn_convs.2.conv.bias - torch.Size([1024]):
The value is the same before and after calling init_weights of EncoderDecoder

neck.fpn_convs.3.conv.weight - torch.Size([1024, 1024, 3, 3]):
XavierInit: gain=1, distribution=uniform, bias=0

neck.fpn_convs.3.conv.bias - torch.Size([1024]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.conv_seg.weight - torch.Size([2, 512, 1, 1]):
NormalInit: mean=0, std=0.01, bias=0

decode_head.conv_seg.bias - torch.Size([2]):
NormalInit: mean=0, std=0.01, bias=0

decode_head.scale_heads.0.0.conv.weight - torch.Size([512, 1024, 3, 3]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.0.0.bn.weight - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.0.0.bn.bias - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.1.0.conv.weight - torch.Size([512, 1024, 3, 3]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.1.0.bn.weight - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.1.0.bn.bias - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.2.0.conv.weight - torch.Size([512, 1024, 3, 3]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.2.0.bn.weight - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.2.0.bn.bias - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.2.2.conv.weight - torch.Size([512, 512, 3, 3]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.2.2.bn.weight - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.2.2.bn.bias - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.3.0.conv.weight - torch.Size([512, 1024, 3, 3]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.3.0.bn.weight - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.3.0.bn.bias - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.3.2.conv.weight - torch.Size([512, 512, 3, 3]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.3.2.bn.weight - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.3.2.bn.bias - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.3.4.conv.weight - torch.Size([512, 512, 3, 3]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.3.4.bn.weight - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder

decode_head.scale_heads.3.4.bn.bias - torch.Size([512]):
The value is the same before and after calling init_weights of EncoderDecoder
2023-06-30 16:39:02,828 - mmseg - INFO - EncoderDecoder(
(backbone): Vanillanet(
(stem1): Sequential(
(0): Conv2d(3, 512, kernel_size=(4, 4), stride=(4, 4))
(1): _BatchNormXd(512, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(stem2): Sequential(
(0): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(512, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
(2): activation(
(bn): _BatchNormXd(512, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
)
(stages): ModuleList(
(0): Block(
(conv1): Sequential(
(0): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(512, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(conv2): Sequential(
(0): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(512, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(pool): Identity()
(act): activation(
(bn): _BatchNormXd(512, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): Block(
(conv1): Sequential(
(0): Conv2d(512, 512, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(512, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(conv2): Sequential(
(0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(1024, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(pool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(act): activation(
(bn): _BatchNormXd(1024, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
)
(2): Block(
(conv1): Sequential(
(0): Conv2d(1024, 1024, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(1024, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(conv2): Sequential(
(0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(pool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(act): activation(
(bn): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
)
(3): Block(
(conv1): Sequential(
(0): Conv2d(2048, 2048, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(conv2): Sequential(
(0): Conv2d(2048, 2048, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(pool): Identity()
(act): activation(
(bn): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
)
(4): Block(
(conv1): Sequential(
(0): Conv2d(2048, 2048, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(conv2): Sequential(
(0): Conv2d(2048, 2048, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(pool): Identity()
(act): activation(
(bn): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
)
(5): Block(
(conv1): Sequential(
(0): Conv2d(2048, 2048, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(conv2): Sequential(
(0): Conv2d(2048, 2048, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(pool): Identity()
(act): activation(
(bn): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
)
(6): Block(
(conv1): Sequential(
(0): Conv2d(2048, 2048, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(conv2): Sequential(
(0): Conv2d(2048, 2048, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(pool): Identity()
(act): activation(
(bn): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
)
(7): Block(
(conv1): Sequential(
(0): Conv2d(2048, 2048, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(conv2): Sequential(
(0): Conv2d(2048, 2048, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(pool): Identity()
(act): activation(
(bn): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
)
(8): Block(
(conv1): Sequential(
(0): Conv2d(2048, 2048, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(conv2): Sequential(
(0): Conv2d(2048, 2048, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(pool): Identity()
(act): activation(
(bn): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
)
(9): Block(
(conv1): Sequential(
(0): Conv2d(2048, 2048, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(2048, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(conv2): Sequential(
(0): Conv2d(2048, 4096, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(4096, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(pool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(act): activation(
(bn): _BatchNormXd(4096, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
)
(10): Block(
(conv1): Sequential(
(0): Conv2d(4096, 4096, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(4096, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(conv2): Sequential(
(0): Conv2d(4096, 4096, kernel_size=(1, 1), stride=(1, 1))
(1): _BatchNormXd(4096, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
(pool): Identity()
(act): activation(
(bn): _BatchNormXd(4096, eps=1e-06, momentum=0.1, affine=True, track_running_stats=True)
)
)
)
)
(neck): FPN(
(lateral_convs): ModuleList(
(0): ConvModule(
(conv): Conv2d(512, 1024, kernel_size=(1, 1), stride=(1, 1))
)
(1): ConvModule(
(conv): Conv2d(1024, 1024, kernel_size=(1, 1), stride=(1, 1))
)
(2): ConvModule(
(conv): Conv2d(2048, 1024, kernel_size=(1, 1), stride=(1, 1))
)
(3): ConvModule(
(conv): Conv2d(4096, 1024, kernel_size=(1, 1), stride=(1, 1))
)
)
(fpn_convs): ModuleList(
(0): ConvModule(
(conv): Conv2d(1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
)
(1): ConvModule(
(conv): Conv2d(1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
)
(2): ConvModule(
(conv): Conv2d(1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
)
(3): ConvModule(
(conv): Conv2d(1024, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
)
)
)
init_cfg={'type': 'Xavier', 'layer': 'Conv2d', 'distribution': 'uniform'}
(decode_head): FPNHead(
input_transform=multiple_select, ignore_index=255, align_corners=False
(loss_decode): CrossEntropyLoss(avg_non_ignore=False)
(conv_seg): Conv2d(512, 2, kernel_size=(1, 1), stride=(1, 1))
(dropout): Dropout2d(p=0.1, inplace=False)
(scale_heads): ModuleList(
(0): Sequential(
(0): ConvModule(
(conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(activate): ReLU(inplace=True)
)
)
(1): Sequential(
(0): ConvModule(
(conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(activate): ReLU(inplace=True)
)
(1): Upsample()
)
(2): Sequential(
(0): ConvModule(
(conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(activate): ReLU(inplace=True)
)
(1): Upsample()
(2): ConvModule(
(conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(activate): ReLU(inplace=True)
)
(3): Upsample()
)
(3): Sequential(
(0): ConvModule(
(conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(activate): ReLU(inplace=True)
)
(1): Upsample()
(2): ConvModule(
(conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(activate): ReLU(inplace=True)
)
(3): Upsample()
(4): ConvModule(
(conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(activate): ReLU(inplace=True)
)
(5): Upsample()
)
)
)
init_cfg={'type': 'Normal', 'std': 0.01, 'override': {'name': 'conv_seg'}}
)
2023-06-30 16:39:02,845 - mmseg - INFO - Loaded 1341 images
2023-06-30 16:39:02,847 - mmseg - INFO - Loaded 178 images
2023-06-30 16:39:03,842 - mmseg - INFO - self.paramwise_cfg is {'decay_rate': 0.6, 'decay_type': 'layer_wise', 'num_layers': 6}
2023-06-30 16:39:03,842 - mmseg - INFO - Build LearningRateDecayOptimizerConstructor layer_wise 0.6 - 8
2023-06-30 16:39:03,842 - mmseg - INFO - set param backbone.stem1.0.weight as id 0
2023-06-30 16:39:03,843 - mmseg - INFO - set param backbone.stem1.0.bias as id 0
2023-06-30 16:39:03,843 - mmseg - INFO - set param backbone.stem1.1.weight as id 0
2023-06-30 16:39:03,843 - mmseg - INFO - set param backbone.stem1.1.bias as id 0
2023-06-30 16:39:03,843 - mmseg - INFO - set param backbone.stem2.0.weight as id 0
2023-06-30 16:39:03,843 - mmseg - INFO - set param backbone.stem2.0.bias as id 0
2023-06-30 16:39:03,843 - mmseg - INFO - set param backbone.stem2.1.weight as id 0
2023-06-30 16:39:03,843 - mmseg - INFO - set param backbone.stem2.1.bias as id 0
2023-06-30 16:39:03,843 - mmseg - INFO - set param backbone.stem2.2.weight as id 0
2023-06-30 16:39:03,843 - mmseg - INFO - set param backbone.stem2.2.bn.weight as id 0
2023-06-30 16:39:03,843 - mmseg - INFO - set param backbone.stem2.2.bn.bias as id 0
2023-06-30 16:39:03,844 - mmseg - INFO - set param backbone.stages.0.conv1.0.weight as id 1
2023-06-30 16:39:03,844 - mmseg - INFO - set param backbone.stages.0.conv1.0.bias as id 1
2023-06-30 16:39:03,844 - mmseg - INFO - set param backbone.stages.0.conv1.1.weight as id 1
2023-06-30 16:39:03,844 - mmseg - INFO - set param backbone.stages.0.conv1.1.bias as id 1
2023-06-30 16:39:03,844 - mmseg - INFO - set param backbone.stages.0.conv2.0.weight as id 1
2023-06-30 16:39:03,844 - mmseg - INFO - set param backbone.stages.0.conv2.0.bias as id 1
2023-06-30 16:39:03,844 - mmseg - INFO - set param backbone.stages.0.conv2.1.weight as id 1
2023-06-30 16:39:03,844 - mmseg - INFO - set param backbone.stages.0.conv2.1.bias as id 1
2023-06-30 16:39:03,844 - mmseg - INFO - set param backbone.stages.0.act.weight as id 1
2023-06-30 16:39:03,845 - mmseg - INFO - set param backbone.stages.0.act.bn.weight as id 1
2023-06-30 16:39:03,845 - mmseg - INFO - set param backbone.stages.0.act.bn.bias as id 1
2023-06-30 16:39:03,845 - mmseg - INFO - set param backbone.stages.1.conv1.0.weight as id 1
2023-06-30 16:39:03,845 - mmseg - INFO - set param backbone.stages.1.conv1.0.bias as id 1
2023-06-30 16:39:03,845 - mmseg - INFO - set param backbone.stages.1.conv1.1.weight as id 1
2023-06-30 16:39:03,845 - mmseg - INFO - set param backbone.stages.1.conv1.1.bias as id 1
2023-06-30 16:39:03,845 - mmseg - INFO - set param backbone.stages.1.conv2.0.weight as id 1
2023-06-30 16:39:03,845 - mmseg - INFO - set param backbone.stages.1.conv2.0.bias as id 1
2023-06-30 16:39:03,845 - mmseg - INFO - set param backbone.stages.1.conv2.1.weight as id 1
2023-06-30 16:39:03,845 - mmseg - INFO - set param backbone.stages.1.conv2.1.bias as id 1
2023-06-30 16:39:03,846 - mmseg - INFO - set param backbone.stages.1.act.weight as id 1
2023-06-30 16:39:03,846 - mmseg - INFO - set param backbone.stages.1.act.bn.weight as id 1
2023-06-30 16:39:03,846 - mmseg - INFO - set param backbone.stages.1.act.bn.bias as id 1
2023-06-30 16:39:03,846 - mmseg - INFO - set param backbone.stages.2.conv1.0.weight as id 2
2023-06-30 16:39:03,846 - mmseg - INFO - set param backbone.stages.2.conv1.0.bias as id 2
2023-06-30 16:39:03,846 - mmseg - INFO - set param backbone.stages.2.conv1.1.weight as id 2
2023-06-30 16:39:03,846 - mmseg - INFO - set param backbone.stages.2.conv1.1.bias as id 2
2023-06-30 16:39:03,846 - mmseg - INFO - set param backbone.stages.2.conv2.0.weight as id 2
2023-06-30 16:39:03,846 - mmseg - INFO - set param backbone.stages.2.conv2.0.bias as id 2
2023-06-30 16:39:03,846 - mmseg - INFO - set param backbone.stages.2.conv2.1.weight as id 2
2023-06-30 16:39:03,847 - mmseg - INFO - set param backbone.stages.2.conv2.1.bias as id 2
2023-06-30 16:39:03,847 - mmseg - INFO - set param backbone.stages.2.act.weight as id 2
2023-06-30 16:39:03,847 - mmseg - INFO - set param backbone.stages.2.act.bn.weight as id 2
2023-06-30 16:39:03,847 - mmseg - INFO - set param backbone.stages.2.act.bn.bias as id 2
2023-06-30 16:39:03,847 - mmseg - INFO - set param backbone.stages.3.conv1.0.weight as id 2
2023-06-30 16:39:03,847 - mmseg - INFO - set param backbone.stages.3.conv1.0.bias as id 2
2023-06-30 16:39:03,847 - mmseg - INFO - set param backbone.stages.3.conv1.1.weight as id 2
2023-06-30 16:39:03,847 - mmseg - INFO - set param backbone.stages.3.conv1.1.bias as id 2
2023-06-30 16:39:03,848 - mmseg - INFO - set param backbone.stages.3.conv2.0.weight as id 2
2023-06-30 16:39:03,848 - mmseg - INFO - set param backbone.stages.3.conv2.0.bias as id 2
2023-06-30 16:39:03,848 - mmseg - INFO - set param backbone.stages.3.conv2.1.weight as id 2
2023-06-30 16:39:03,848 - mmseg - INFO - set param backbone.stages.3.conv2.1.bias as id 2
2023-06-30 16:39:03,848 - mmseg - INFO - set param backbone.stages.3.act.weight as id 2
2023-06-30 16:39:03,848 - mmseg - INFO - set param backbone.stages.3.act.bn.weight as id 2
2023-06-30 16:39:03,848 - mmseg - INFO - set param backbone.stages.3.act.bn.bias as id 2
2023-06-30 16:39:03,848 - mmseg - INFO - set param backbone.stages.4.conv1.0.weight as id 3
2023-06-30 16:39:03,848 - mmseg - INFO - set param backbone.stages.4.conv1.0.bias as id 3
2023-06-30 16:39:03,848 - mmseg - INFO - set param backbone.stages.4.conv1.1.weight as id 3
2023-06-30 16:39:03,849 - mmseg - INFO - set param backbone.stages.4.conv1.1.bias as id 3
2023-06-30 16:39:03,849 - mmseg - INFO - set param backbone.stages.4.conv2.0.weight as id 3
2023-06-30 16:39:03,849 - mmseg - INFO - set param backbone.stages.4.conv2.0.bias as id 3
2023-06-30 16:39:03,849 - mmseg - INFO - set param backbone.stages.4.conv2.1.weight as id 3
2023-06-30 16:39:03,849 - mmseg - INFO - set param backbone.stages.4.conv2.1.bias as id 3
2023-06-30 16:39:03,849 - mmseg - INFO - set param backbone.stages.4.act.weight as id 3
2023-06-30 16:39:03,849 - mmseg - INFO - set param backbone.stages.4.act.bn.weight as id 3
2023-06-30 16:39:03,849 - mmseg - INFO - set param backbone.stages.4.act.bn.bias as id 3
2023-06-30 16:39:03,849 - mmseg - INFO - set param backbone.stages.5.conv1.0.weight as id 3
2023-06-30 16:39:03,850 - mmseg - INFO - set param backbone.stages.5.conv1.0.bias as id 3
2023-06-30 16:39:03,850 - mmseg - INFO - set param backbone.stages.5.conv1.1.weight as id 3
2023-06-30 16:39:03,850 - mmseg - INFO - set param backbone.stages.5.conv1.1.bias as id 3
2023-06-30 16:39:03,850 - mmseg - INFO - set param backbone.stages.5.conv2.0.weight as id 3
2023-06-30 16:39:03,850 - mmseg - INFO - set param backbone.stages.5.conv2.0.bias as id 3
2023-06-30 16:39:03,850 - mmseg - INFO - set param backbone.stages.5.conv2.1.weight as id 3
2023-06-30 16:39:03,850 - mmseg - INFO - set param backbone.stages.5.conv2.1.bias as id 3
2023-06-30 16:39:03,850 - mmseg - INFO - set param backbone.stages.5.act.weight as id 3
2023-06-30 16:39:03,850 - mmseg - INFO - set param backbone.stages.5.act.bn.weight as id 3
2023-06-30 16:39:03,850 - mmseg - INFO - set param backbone.stages.5.act.bn.bias as id 3
2023-06-30 16:39:03,851 - mmseg - INFO - set param backbone.stages.6.conv1.0.weight as id 4
2023-06-30 16:39:03,851 - mmseg - INFO - set param backbone.stages.6.conv1.0.bias as id 4
2023-06-30 16:39:03,851 - mmseg - INFO - set param backbone.stages.6.conv1.1.weight as id 4
2023-06-30 16:39:03,851 - mmseg - INFO - set param backbone.stages.6.conv1.1.bias as id 4
2023-06-30 16:39:03,851 - mmseg - INFO - set param backbone.stages.6.conv2.0.weight as id 4
2023-06-30 16:39:03,851 - mmseg - INFO - set param backbone.stages.6.conv2.0.bias as id 4
2023-06-30 16:39:03,851 - mmseg - INFO - set param backbone.stages.6.conv2.1.weight as id 4
2023-06-30 16:39:03,851 - mmseg - INFO - set param backbone.stages.6.conv2.1.bias as id 4
2023-06-30 16:39:03,851 - mmseg - INFO - set param backbone.stages.6.act.weight as id 4
2023-06-30 16:39:03,851 - mmseg - INFO - set param backbone.stages.6.act.bn.weight as id 4
2023-06-30 16:39:03,852 - mmseg - INFO - set param backbone.stages.6.act.bn.bias as id 4
2023-06-30 16:39:03,862 - mmseg - INFO - set param backbone.stages.7.conv1.0.weight as id 4
2023-06-30 16:39:03,862 - mmseg - INFO - set param backbone.stages.7.conv1.0.bias as id 4
2023-06-30 16:39:03,862 - mmseg - INFO - set param backbone.stages.7.conv1.1.weight as id 4
2023-06-30 16:39:03,862 - mmseg - INFO - set param backbone.stages.7.conv1.1.bias as id 4
2023-06-30 16:39:03,862 - mmseg - INFO - set param backbone.stages.7.conv2.0.weight as id 4
2023-06-30 16:39:03,862 - mmseg - INFO - set param backbone.stages.7.conv2.0.bias as id 4
2023-06-30 16:39:03,863 - mmseg - INFO - set param backbone.stages.7.conv2.1.weight as id 4
2023-06-30 16:39:03,863 - mmseg - INFO - set param backbone.stages.7.conv2.1.bias as id 4
2023-06-30 16:39:03,863 - mmseg - INFO - set param backbone.stages.7.act.weight as id 4
2023-06-30 16:39:03,863 - mmseg - INFO - set param backbone.stages.7.act.bn.weight as id 4
2023-06-30 16:39:03,863 - mmseg - INFO - set param backbone.stages.7.act.bn.bias as id 4
2023-06-30 16:39:03,863 - mmseg - INFO - set param backbone.stages.8.conv1.0.weight as id 5
2023-06-30 16:39:03,863 - mmseg - INFO - set param backbone.stages.8.conv1.0.bias as id 5
2023-06-30 16:39:03,863 - mmseg - INFO - set param backbone.stages.8.conv1.1.weight as id 5
2023-06-30 16:39:03,863 - mmseg - INFO - set param backbone.stages.8.conv1.1.bias as id 5
2023-06-30 16:39:03,863 - mmseg - INFO - set param backbone.stages.8.conv2.0.weight as id 5
2023-06-30 16:39:03,864 - mmseg - INFO - set param backbone.stages.8.conv2.0.bias as id 5
2023-06-30 16:39:03,864 - mmseg - INFO - set param backbone.stages.8.conv2.1.weight as id 5
2023-06-30 16:39:03,864 - mmseg - INFO - set param backbone.stages.8.conv2.1.bias as id 5
2023-06-30 16:39:03,864 - mmseg - INFO - set param backbone.stages.8.act.weight as id 5
2023-06-30 16:39:03,864 - mmseg - INFO - set param backbone.stages.8.act.bn.weight as id 5
2023-06-30 16:39:03,864 - mmseg - INFO - set param backbone.stages.8.act.bn.bias as id 5
2023-06-30 16:39:03,864 - mmseg - INFO - set param backbone.stages.9.conv1.0.weight as id 5
2023-06-30 16:39:03,864 - mmseg - INFO - set param backbone.stages.9.conv1.0.bias as id 5
2023-06-30 16:39:03,864 - mmseg - INFO - set param backbone.stages.9.conv1.1.weight as id 5
2023-06-30 16:39:03,864 - mmseg - INFO - set param backbone.stages.9.conv1.1.bias as id 5
2023-06-30 16:39:03,865 - mmseg - INFO - set param backbone.stages.9.conv2.0.weight as id 5
2023-06-30 16:39:03,865 - mmseg - INFO - set param backbone.stages.9.conv2.0.bias as id 5
2023-06-30 16:39:03,865 - mmseg - INFO - set param backbone.stages.9.conv2.1.weight as id 5
2023-06-30 16:39:03,865 - mmseg - INFO - set param backbone.stages.9.conv2.1.bias as id 5
2023-06-30 16:39:03,865 - mmseg - INFO - set param backbone.stages.9.act.weight as id 5
2023-06-30 16:39:03,865 - mmseg - INFO - set param backbone.stages.9.act.bn.weight as id 5
2023-06-30 16:39:03,865 - mmseg - INFO - set param backbone.stages.9.act.bn.bias as id 5
2023-06-30 16:39:03,865 - mmseg - INFO - set param backbone.stages.10.conv1.0.weight as id 6
2023-06-30 16:39:03,865 - mmseg - INFO - set param backbone.stages.10.conv1.0.bias as id 6
2023-06-30 16:39:03,866 - mmseg - INFO - set param backbone.stages.10.conv1.1.weight as id 6
2023-06-30 16:39:03,866 - mmseg - INFO - set param backbone.stages.10.conv1.1.bias as id 6
2023-06-30 16:39:03,866 - mmseg - INFO - set param backbone.stages.10.conv2.0.weight as id 6
2023-06-30 16:39:03,866 - mmseg - INFO - set param backbone.stages.10.conv2.0.bias as id 6
2023-06-30 16:39:03,866 - mmseg - INFO - set param backbone.stages.10.conv2.1.weight as id 6
2023-06-30 16:39:03,866 - mmseg - INFO - set param backbone.stages.10.conv2.1.bias as id 6
2023-06-30 16:39:03,866 - mmseg - INFO - set param backbone.stages.10.act.weight as id 6
2023-06-30 16:39:03,866 - mmseg - INFO - set param backbone.stages.10.act.bn.weight as id 6
2023-06-30 16:39:03,866 - mmseg - INFO - set param backbone.stages.10.act.bn.bias as id 6
2023-06-30 16:39:03,866 - mmseg - INFO - set param neck.lateral_convs.0.conv.weight as id 7
2023-06-30 16:39:03,878 - mmseg - INFO - set param neck.lateral_convs.0.conv.bias as id 7
2023-06-30 16:39:03,878 - mmseg - INFO - set param neck.lateral_convs.1.conv.weight as id 7
2023-06-30 16:39:03,878 - mmseg - INFO - set param neck.lateral_convs.1.conv.bias as id 7
2023-06-30 16:39:03,878 - mmseg - INFO - set param neck.lateral_convs.2.conv.weight as id 7
2023-06-30 16:39:03,878 - mmseg - INFO - set param neck.lateral_convs.2.conv.bias as id 7
2023-06-30 16:39:03,878 - mmseg - INFO - set param neck.lateral_convs.3.conv.weight as id 7
2023-06-30 16:39:03,879 - mmseg - INFO - set param neck.lateral_convs.3.conv.bias as id 7
2023-06-30 16:39:03,879 - mmseg - INFO - set param neck.fpn_convs.0.conv.weight as id 7
2023-06-30 16:39:03,879 - mmseg - INFO - set param neck.fpn_convs.0.conv.bias as id 7
2023-06-30 16:39:03,879 - mmseg - INFO - set param neck.fpn_convs.1.conv.weight as id 7
2023-06-30 16:39:03,879 - mmseg - INFO - set param neck.fpn_convs.1.conv.bias as id 7
2023-06-30 16:39:03,879 - mmseg - INFO - set param neck.fpn_convs.2.conv.weight as id 7
2023-06-30 16:39:03,879 - mmseg - INFO - set param neck.fpn_convs.2.conv.bias as id 7
2023-06-30 16:39:03,879 - mmseg - INFO - set param neck.fpn_convs.3.conv.weight as id 7
2023-06-30 16:39:03,879 - mmseg - INFO - set param neck.fpn_convs.3.conv.bias as id 7
2023-06-30 16:39:03,879 - mmseg - INFO - set param decode_head.conv_seg.weight as id 7
2023-06-30 16:39:03,880 - mmseg - INFO - set param decode_head.conv_seg.bias as id 7
2023-06-30 16:39:03,880 - mmseg - INFO - set param decode_head.scale_heads.0.0.conv.weight as id 7
2023-06-30 16:39:03,880 - mmseg - INFO - set param decode_head.scale_heads.0.0.bn.weight as id 7
2023-06-30 16:39:03,880 - mmseg - INFO - set param decode_head.scale_heads.0.0.bn.bias as id 7
2023-06-30 16:39:03,880 - mmseg - INFO - set param decode_head.scale_heads.1.0.conv.weight as id 7
2023-06-30 16:39:03,880 - mmseg - INFO - set param decode_head.scale_heads.1.0.bn.weight as id 7
2023-06-30 16:39:03,880 - mmseg - INFO - set param decode_head.scale_heads.1.0.bn.bias as id 7
2023-06-30 16:39:03,880 - mmseg - INFO - set param decode_head.scale_heads.2.0.conv.weight as id 7
2023-06-30 16:39:03,880 - mmseg - INFO - set param decode_head.scale_heads.2.0.bn.weight as id 7
2023-06-30 16:39:03,880 - mmseg - INFO - set param decode_head.scale_heads.2.0.bn.bias as id 7
2023-06-30 16:39:03,881 - mmseg - INFO - set param decode_head.scale_heads.2.2.conv.weight as id 7
2023-06-30 16:39:03,881 - mmseg - INFO - set param decode_head.scale_heads.2.2.bn.weight as id 7
2023-06-30 16:39:03,881 - mmseg - INFO - set param decode_head.scale_heads.2.2.bn.bias as id 7
2023-06-30 16:39:03,881 - mmseg - INFO - set param decode_head.scale_heads.3.0.conv.weight as id 7
2023-06-30 16:39:03,881 - mmseg - INFO - set param decode_head.scale_heads.3.0.bn.weight as id 7
2023-06-30 16:39:03,881 - mmseg - INFO - set param decode_head.scale_heads.3.0.bn.bias as id 7
2023-06-30 16:39:03,881 - mmseg - INFO - set param decode_head.scale_heads.3.2.conv.weight as id 7
2023-06-30 16:39:03,881 - mmseg - INFO - set param decode_head.scale_heads.3.2.bn.weight as id 7
2023-06-30 16:39:03,881 - mmseg - INFO - set param decode_head.scale_heads.3.2.bn.bias as id 7
2023-06-30 16:39:03,882 - mmseg - INFO - set param decode_head.scale_heads.3.4.conv.weight as id 7
2023-06-30 16:39:03,882 - mmseg - INFO - set param decode_head.scale_heads.3.4.bn.weight as id 7
2023-06-30 16:39:03,882 - mmseg - INFO - set param decode_head.scale_heads.3.4.bn.bias as id 7
2023-06-30 16:39:03,882 - mmseg - INFO - Param groups = {
"layer_0_decay": {
"param_names": [
"backbone.stem1.0.weight",
"backbone.stem2.0.weight",
"backbone.stem2.2.weight"
],
"lr_scale": 0.027993599999999993,
"lr": 3.639167999999999e-06,
"weight_decay": 0.05
},
"layer_0_no_decay": {
"param_names": [
"backbone.stem1.0.bias",
"backbone.stem1.1.weight",
"backbone.stem1.1.bias",
"backbone.stem2.0.bias",
"backbone.stem2.1.weight",
"backbone.stem2.1.bias",
"backbone.stem2.2.bn.weight",
"backbone.stem2.2.bn.bias"
],
"lr_scale": 0.027993599999999993,
"lr": 3.639167999999999e-06,
"weight_decay": 0.0
},
"layer_1_decay": {
"param_names": [
"backbone.stages.0.conv1.0.weight",
"backbone.stages.0.conv2.0.weight",
"backbone.stages.0.act.weight",
"backbone.stages.1.conv1.0.weight",
"backbone.stages.1.conv2.0.weight",
"backbone.stages.1.act.weight"
],
"lr_scale": 0.04665599999999999,
"lr": 6.065279999999998e-06,
"weight_decay": 0.05
},
"layer_1_no_decay": {
"param_names": [
"backbone.stages.0.conv1.0.bias",
"backbone.stages.0.conv1.1.weight",
"backbone.stages.0.conv1.1.bias",
"backbone.stages.0.conv2.0.bias",
"backbone.stages.0.conv2.1.weight",
"backbone.stages.0.conv2.1.bias",
"backbone.stages.0.act.bn.weight",
"backbone.stages.0.act.bn.bias",
"backbone.stages.1.conv1.0.bias",
"backbone.stages.1.conv1.1.weight",
"backbone.stages.1.conv1.1.bias",
"backbone.stages.1.conv2.0.bias",
"backbone.stages.1.conv2.1.weight",
"backbone.stages.1.conv2.1.bias",
"backbone.stages.1.act.bn.weight",
"backbone.stages.1.act.bn.bias"
],
"lr_scale": 0.04665599999999999,
"lr": 6.065279999999998e-06,
"weight_decay": 0.0
},
"layer_2_decay": {
"param_names": [
"backbone.stages.2.conv1.0.weight",
"backbone.stages.2.conv2.0.weight",
"backbone.stages.2.act.weight",
"backbone.stages.3.conv1.0.weight",
"backbone.stages.3.conv2.0.weight",
"backbone.stages.3.act.weight"
],
"lr_scale": 0.07775999999999998,
"lr": 1.0108799999999997e-05,
"weight_decay": 0.05
},
"layer_2_no_decay": {
"param_names": [
"backbone.stages.2.conv1.0.bias",
"backbone.stages.2.conv1.1.weight",
"backbone.stages.2.conv1.1.bias",
"backbone.stages.2.conv2.0.bias",
"backbone.stages.2.conv2.1.weight",
"backbone.stages.2.conv2.1.bias",
"backbone.stages.2.act.bn.weight",
"backbone.stages.2.act.bn.bias",
"backbone.stages.3.conv1.0.bias",
"backbone.stages.3.conv1.1.weight",
"backbone.stages.3.conv1.1.bias",
"backbone.stages.3.conv2.0.bias",
"backbone.stages.3.conv2.1.weight",
"backbone.stages.3.conv2.1.bias",
"backbone.stages.3.act.bn.weight",
"backbone.stages.3.act.bn.bias"
],
"lr_scale": 0.07775999999999998,
"lr": 1.0108799999999997e-05,
"weight_decay": 0.0
},
"layer_3_decay": {
"param_names": [
"backbone.stages.4.conv1.0.weight",
"backbone.stages.4.conv2.0.weight",
"backbone.stages.4.act.weight",
"backbone.stages.5.conv1.0.weight",
"backbone.stages.5.conv2.0.weight",
"backbone.stages.5.act.weight"
],
"lr_scale": 0.1296,
"lr": 1.6848e-05,
"weight_decay": 0.05
},
"layer_3_no_decay": {
"param_names": [
"backbone.stages.4.conv1.0.bias",
"backbone.stages.4.conv1.1.weight",
"backbone.stages.4.conv1.1.bias",
"backbone.stages.4.conv2.0.bias",
"backbone.stages.4.conv2.1.weight",
"backbone.stages.4.conv2.1.bias",
"backbone.stages.4.act.bn.weight",
"backbone.stages.4.act.bn.bias",
"backbone.stages.5.conv1.0.bias",
"backbone.stages.5.conv1.1.weight",
"backbone.stages.5.conv1.1.bias",
"backbone.stages.5.conv2.0.bias",
"backbone.stages.5.conv2.1.weight",
"backbone.stages.5.conv2.1.bias",
"backbone.stages.5.act.bn.weight",
"backbone.stages.5.act.bn.bias"
],
"lr_scale": 0.1296,
"lr": 1.6848e-05,
"weight_decay": 0.0
},
"layer_4_decay": {
"param_names": [
"backbone.stages.6.conv1.0.weight",
"backbone.stages.6.conv2.0.weight",
"backbone.stages.6.act.weight",
"backbone.stages.7.conv1.0.weight",
"backbone.stages.7.conv2.0.weight",
"backbone.stages.7.act.weight"
],
"lr_scale": 0.21599999999999997,
"lr": 2.8079999999999992e-05,
"weight_decay": 0.05
},
"layer_4_no_decay": {
"param_names": [
"backbone.stages.6.conv1.0.bias",
"backbone.stages.6.conv1.1.weight",
"backbone.stages.6.conv1.1.bias",
"backbone.stages.6.conv2.0.bias",
"backbone.stages.6.conv2.1.weight",
"backbone.stages.6.conv2.1.bias",
"backbone.stages.6.act.bn.weight",
"backbone.stages.6.act.bn.bias",
"backbone.stages.7.conv1.0.bias",
"backbone.stages.7.conv1.1.weight",
"backbone.stages.7.conv1.1.bias",
"backbone.stages.7.conv2.0.bias",
"backbone.stages.7.conv2.1.weight",
"backbone.stages.7.conv2.1.bias",
"backbone.stages.7.act.bn.weight",
"backbone.stages.7.act.bn.bias"
],
"lr_scale": 0.21599999999999997,
"lr": 2.8079999999999992e-05,
"weight_decay": 0.0
},
"layer_5_decay": {
"param_names": [
"backbone.stages.8.conv1.0.weight",
"backbone.stages.8.conv2.0.weight",
"backbone.stages.8.act.weight",
"backbone.stages.9.conv1.0.weight",
"backbone.stages.9.conv2.0.weight",
"backbone.stages.9.act.weight"
],
"lr_scale": 0.36,
"lr": 4.679999999999999e-05,
"weight_decay": 0.05
},
"layer_5_no_decay": {
"param_names": [
"backbone.stages.8.conv1.0.bias",
"backbone.stages.8.conv1.1.weight",
"backbone.stages.8.conv1.1.bias",
"backbone.stages.8.conv2.0.bias",
"backbone.stages.8.conv2.1.weight",
"backbone.stages.8.conv2.1.bias",
"backbone.stages.8.act.bn.weight",
"backbone.stages.8.act.bn.bias",
"backbone.stages.9.conv1.0.bias",
"backbone.stages.9.conv1.1.weight",
"backbone.stages.9.conv1.1.bias",
"backbone.stages.9.conv2.0.bias",
"backbone.stages.9.conv2.1.weight",
"backbone.stages.9.conv2.1.bias",
"backbone.stages.9.act.bn.weight",
"backbone.stages.9.act.bn.bias"
],
"lr_scale": 0.36,
"lr": 4.679999999999999e-05,
"weight_decay": 0.0
},
"layer_6_decay": {
"param_names": [
"backbone.stages.10.conv1.0.weight",
"backbone.stages.10.conv2.0.weight",
"backbone.stages.10.act.weight"
],
"lr_scale": 0.6,
"lr": 7.799999999999999e-05,
"weight_decay": 0.05
},
"layer_6_no_decay": {
"param_names": [
"backbone.stages.10.conv1.0.bias",
"backbone.stages.10.conv1.1.weight",
"backbone.stages.10.conv1.1.bias",
"backbone.stages.10.conv2.0.bias",
"backbone.stages.10.conv2.1.weight",
"backbone.stages.10.conv2.1.bias",
"backbone.stages.10.act.bn.weight",
"backbone.stages.10.act.bn.bias"
],
"lr_scale": 0.6,
"lr": 7.799999999999999e-05,
"weight_decay": 0.0
},
"layer_7_decay": {
"param_names": [
"neck.lateral_convs.0.conv.weight",
"neck.lateral_convs.1.conv.weight",
"neck.lateral_convs.2.conv.weight",
"neck.lateral_convs.3.conv.weight",
"neck.fpn_convs.0.conv.weight",
"neck.fpn_convs.1.conv.weight",
"neck.fpn_convs.2.conv.weight",
"neck.fpn_convs.3.conv.weight",
"decode_head.conv_seg.weight",
"decode_head.scale_heads.0.0.conv.weight",
"decode_head.scale_heads.1.0.conv.weight",
"decode_head.scale_heads.2.0.conv.weight",
"decode_head.scale_heads.2.2.conv.weight",
"decode_head.scale_heads.3.0.conv.weight",
"decode_head.scale_heads.3.2.conv.weight",
"decode_head.scale_heads.3.4.conv.weight"
],
"lr_scale": 1.0,
"lr": 0.00013,
"weight_decay": 0.05
},
"layer_7_no_decay": {
"param_names": [
"neck.lateral_convs.0.conv.bias",
"neck.lateral_convs.1.conv.bias",
"neck.lateral_convs.2.conv.bias",
"neck.lateral_convs.3.conv.bias",
"neck.fpn_convs.0.conv.bias",
"neck.fpn_convs.1.conv.bias",
"neck.fpn_convs.2.conv.bias",
"neck.fpn_convs.3.conv.bias",
"decode_head.conv_seg.bias",
"decode_head.scale_heads.0.0.bn.weight",
"decode_head.scale_heads.0.0.bn.bias",
"decode_head.scale_heads.1.0.bn.weight",
"decode_head.scale_heads.1.0.bn.bias",
"decode_head.scale_heads.2.0.bn.weight",
"decode_head.scale_heads.2.0.bn.bias",
"decode_head.scale_heads.2.2.bn.weight",
"decode_head.scale_heads.2.2.bn.bias",
"decode_head.scale_heads.3.0.bn.weight",
"decode_head.scale_heads.3.0.bn.bias",
"decode_head.scale_heads.3.2.bn.weight",
"decode_head.scale_heads.3.2.bn.bias",
"decode_head.scale_heads.3.4.bn.weight",
"decode_head.scale_heads.3.4.bn.bias"
],
"lr_scale": 1.0,
"lr": 0.00013,
"weight_decay": 0.0
}
}
2023-06-30 16:39:03,896 - mmseg - INFO - Loaded 178 images
2023-06-30 16:39:03,897 - mmseg - INFO - Start running, host: admin@DESKTOP-DS4MTUV, work_dir: E:\zbl\mmsegmentation-0.23.0\tools\logs_vanillanet\13
2023-06-30 16:39:03,897 - mmseg - INFO - Hooks will be executed in the following order:
before_run:
(VERY_HIGH ) StepLrUpdaterHook
(NORMAL ) CheckpointHook
(LOW ) EvalHook
(VERY_LOW ) TextLoggerHook

before_train_epoch:
(VERY_HIGH ) StepLrUpdaterHook
(LOW ) IterTimerHook
(LOW ) EvalHook
(VERY_LOW ) TextLoggerHook

before_train_iter:
(VERY_HIGH ) StepLrUpdaterHook
(LOW ) IterTimerHook
(LOW ) EvalHook

after_train_iter:
(ABOVE_NORMAL) OptimizerHook
(NORMAL ) CheckpointHook
(LOW ) IterTimerHook
(LOW ) EvalHook
(VERY_LOW ) TextLoggerHook

after_train_epoch:
(NORMAL ) CheckpointHook
(LOW ) EvalHook
(VERY_LOW ) TextLoggerHook

before_val_epoch:
(LOW ) IterTimerHook
(VERY_LOW ) TextLoggerHook

before_val_iter:
(LOW ) IterTimerHook

after_val_iter:
(LOW ) IterTimerHook

after_val_epoch:
(VERY_LOW ) TextLoggerHook

after_run:
(VERY_LOW ) TextLoggerHook

2023-06-30 16:39:03,898 - mmseg - INFO - workflow: [('train', 1), ('val', 1)], max: 20000 iters
2023-06-30 16:39:03,898 - mmseg - INFO - Checkpoints will be saved to E:\zbl\mmsegmentation-0.23.0\tools\logs_vanillanet\13 by HardDiskBackend.
2023-06-30 16:40:00,779 - mmseg - INFO - Iter [20/20000] lr: 1.418e-07, eta: 5:18:40, time: 0.957, data_time: 0.008, memory: 21203, decode.loss_ce: 0.5422, decode.acc_seg: 39.8719, loss: 0.5422
2023-06-30 16:40:40,739 - mmseg - INFO - Iter [40/20000] lr: 2.872e-07, eta: 5:19:25, time: 0.963, data_time: 0.007, memory: 21203, decode.loss_ce: 0.7923, decode.acc_seg: 47.3142, loss: 0.7923
2023-06-30 16:41:21,888 - mmseg - INFO - Iter [60/20000] lr: 4.326e-07, eta: 5:20:22, time: 0.972, data_time: 0.009, memory: 21203, decode.loss_ce: 0.8685, decode.acc_seg: 57.3759, loss: 0.8685

[ggjy]

你可以参考这个config的写法https://github.com/huawei-noah/VanillaNet/blob/main/object_detection/configs/vanillanet/mask_rcnn_vanillanet_13_mstrain_480-1024_adamw.py#L26，把init_cfg=dict(type='Pretrained', checkpoint=checkpoint_file)) 放到 backbone的dict中。

[zbl929]

非常感谢您之前的回复！您的网络效果很不错！
想问一句，将预训练权重放在 backbone的dict与其他加载预训练的方式有什么区别吗？
为什么放在dict里就好使了？
期待回复！

[ggjy]

没有区别，只是我用的mmdet的版本他们把接口放在了dict里面，我只是调用他们的接口，当然你也可以自己写一个去load ckpt，都是没有区别的。