mirrors/mmsegmentation
1
0
Fork 0
mirror of https://github.com/open-mmlab/mmsegmentation.git synced 2025-06-03 22:03:48 +08:00
Code Issues Projects Releases Wiki Activity

[Refactor] Refacor default_hooks and train & val & test loops in configs

Browse Source
This commit is contained in:
limengzhang.vendor 2022-06-08 06:28:35 +00:00 committed by zhengmiao
parent 80bb004bbb
commit c84a58b7b5
49 changed files with 338 additions and 101 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
configs/_base_/default_runtime.py
View File

@ -1,14 +1,9 @@
# yapf:disable
log_config = dict(
interval=50,
hooks=[
dict(type='TextLoggerHook', by_epoch=False),
# dict(type='TensorboardLoggerHook')
])
# yapf:enable
dist_params = dict(backend='nccl')
default_scope = 'mmseg'
env_cfg = dict(
cudnn_benchmark=True,
mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0),
dist_cfg=dict(backend='nccl'),
)
log_level = 'INFO'
load_from = None
resume_from = None
workflow = [('train', 1)]
cudnn_benchmark = True
resume = False

18
configs/_base_/schedules/schedule_160k.py
View File

@ -1,9 +1,19 @@
# optimizer
optimizer = dict(type='SGD', lr=0.01, momentum=0.9, weight_decay=0.0005)
optimizer_config = dict()
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
# learning policy
lr_config = dict(policy='poly', power=0.9, min_lr=1e-4, by_epoch=False)
# runtime settings
runner = dict(type='IterBasedRunner', max_iters=160000)
checkpoint_config = dict(by_epoch=False, interval=16000)
# training schedule for 160k
train_cfg = dict(
type='IterBasedTrainLoop', max_iters=160000, val_interval=16000)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')
evaluation = dict(interval=16000, metric='mIoU', pre_eval=True)
default_hooks = dict(
optimizer=dict(type='OptimizerHook', grad_clip=None),
timer=dict(type='IterTimerHook'),
logger=dict(type='LoggerHook', interval=50),
param_scheduler=dict(type='ParamSchedulerHook'),
checkpoint=dict(type='CheckpointHook', by_epoch=False, interval=16000),
sampler_seed=dict(type='DistSamplerSeedHook'),
)

17
configs/_base_/schedules/schedule_20k.py
View File

@ -1,9 +1,18 @@
# optimizer
optimizer = dict(type='SGD', lr=0.01, momentum=0.9, weight_decay=0.0005)
optimizer_config = dict()
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
# learning policy
lr_config = dict(policy='poly', power=0.9, min_lr=1e-4, by_epoch=False)
# runtime settings
runner = dict(type='IterBasedRunner', max_iters=20000)
checkpoint_config = dict(by_epoch=False, interval=2000)
# training schedule for 20k
train_cfg = dict(type='IterBasedTrainLoop', max_iters=20000, val_interval=2000)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')
evaluation = dict(interval=2000, metric='mIoU', pre_eval=True)
default_hooks = dict(
optimizer=dict(type='OptimizerHook', grad_clip=None),
timer=dict(type='IterTimerHook'),
logger=dict(type='LoggerHook', interval=50),
param_scheduler=dict(type='ParamSchedulerHook'),
checkpoint=dict(type='CheckpointHook', by_epoch=False, interval=2000),
sampler_seed=dict(type='DistSamplerSeedHook'),
)

18
configs/_base_/schedules/schedule_320k.py
View File

@ -1,9 +1,19 @@
# optimizer
optimizer = dict(type='SGD', lr=0.01, momentum=0.9, weight_decay=0.0005)
optimizer_config = dict()
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
# learning policy
lr_config = dict(policy='poly', power=0.9, min_lr=1e-4, by_epoch=False)
# runtime settings
runner = dict(type='IterBasedRunner', max_iters=320000)
checkpoint_config = dict(by_epoch=False, interval=32000)
# training schedule for 320k
train_cfg = dict(
type='IterBasedTrainLoop', max_iters=320000, val_interval=32000)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')
evaluation = dict(interval=32000, metric='mIoU')
default_hooks = dict(
optimizer=dict(type='OptimizerHook', grad_clip=None),
timer=dict(type='IterTimerHook'),
logger=dict(type='LoggerHook', interval=50),
param_scheduler=dict(type='ParamSchedulerHook'),
checkpoint=dict(type='CheckpointHook', by_epoch=False, interval=32000),
sampler_seed=dict(type='DistSamplerSeedHook'),
)

17
configs/_base_/schedules/schedule_40k.py
View File

@ -1,9 +1,18 @@
# optimizer
optimizer = dict(type='SGD', lr=0.01, momentum=0.9, weight_decay=0.0005)
optimizer_config = dict()
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
# learning policy
lr_config = dict(policy='poly', power=0.9, min_lr=1e-4, by_epoch=False)
# runtime settings
runner = dict(type='IterBasedRunner', max_iters=40000)
checkpoint_config = dict(by_epoch=False, interval=4000)
# training schedule for 40k
train_cfg = dict(type='IterBasedTrainLoop', max_iters=40000, val_interval=4000)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')
evaluation = dict(interval=4000, metric='mIoU', pre_eval=True)
default_hooks = dict(
optimizer=dict(type='OptimizerHook', grad_clip=None),
timer=dict(type='IterTimerHook'),
logger=dict(type='LoggerHook', interval=50),
param_scheduler=dict(type='ParamSchedulerHook'),
checkpoint=dict(type='CheckpointHook', by_epoch=False, interval=4000),
sampler_seed=dict(type='DistSamplerSeedHook'),
)

17
configs/_base_/schedules/schedule_80k.py
View File

@ -1,9 +1,18 @@
# optimizer
optimizer = dict(type='SGD', lr=0.01, momentum=0.9, weight_decay=0.0005)
optimizer_config = dict()
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
# learning policy
lr_config = dict(policy='poly', power=0.9, min_lr=1e-4, by_epoch=False)
# runtime settings
runner = dict(type='IterBasedRunner', max_iters=80000)
checkpoint_config = dict(by_epoch=False, interval=8000)
# training schedule for 80k
train_cfg = dict(type='IterBasedTrainLoop', max_iters=80000, val_interval=8000)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')
evaluation = dict(interval=8000, metric='mIoU', pre_eval=True)
default_hooks = dict(
optimizer=dict(type='OptimizerHook', grad_clip=None),
timer=dict(type='IterTimerHook'),
logger=dict(type='LoggerHook', interval=50),
param_scheduler=dict(type='ParamSchedulerHook'),
checkpoint=dict(type='CheckpointHook', by_epoch=False, interval=8000),
sampler_seed=dict(type='DistSamplerSeedHook'),
)

6
configs/beit/upernet_beit-base_8x2_640x640_160k_ade20k.py
View File

@ -12,7 +12,11 @@ optimizer = dict(
type='AdamW',
lr=3e-5,
betas=(0.9, 0.999),
weight_decay=0.05,
weight_decay=0.05)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
constructor='LayerDecayOptimizerConstructor',
paramwise_cfg=dict(num_layers=12, layer_decay_rate=0.9))

7
configs/beit/upernet_beit-large_fp16_8x1_640x640_160k_ade20k.py
View File

@ -26,10 +26,13 @@ optimizer = dict(
type='AdamW',
lr=2e-5,
betas=(0.9, 0.999),
weight_decay=0.05,
weight_decay=0.05)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
constructor='LayerDecayOptimizerConstructor',
paramwise_cfg=dict(num_layers=24, layer_decay_rate=0.95))
lr_config = dict(
_delete_=True,
policy='poly',

6
configs/bisenetv1/bisenetv1_r101-d32_lr5e-3_4x4_512x512_160k_coco-stuff164k.py
View File

@ -15,4 +15,8 @@ model = dict(
dict(in_channels=512, channels=256, num_classes=171),
])
lr_config = dict(warmup='linear', warmup_iters=1000)
optimizer = dict(lr=0.005)
optimizer = dict(type='SGD', lr=0.005, momentum=0.9, weight_decay=0.0005)
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
train_dataloader = dict(batch_size=4, num_workers=4)
val_dataloader = dict(batch_size=4, num_workers=4)
test_dataloader = val_dataloader

3
configs/bisenetv1/bisenetv1_r18-d32_4x4_1024x1024_160k_cityscapes.py
View File

@ -4,7 +4,8 @@ _base_ = [
'../_base_/default_runtime.py', '../_base_/schedules/schedule_160k.py'
]
lr_config = dict(warmup='linear', warmup_iters=1000)
optimizer = dict(lr=0.025)
optimizer = dict(type='SGD', lr=0.025, momentum=0.9, weight_decay=0.0005)
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
train_dataloader = dict(batch_size=4, num_workers=4)
val_dataloader = dict(batch_size=4, num_workers=4)
test_dataloader = val_dataloader

3
configs/bisenetv1/bisenetv1_r18-d32_in1k-pre_4x4_1024x1024_160k_cityscapes.py
View File

@ -9,7 +9,8 @@ model = dict(
init_cfg=dict(
type='Pretrained', checkpoint='open-mmlab://resnet18_v1c'))))
lr_config = dict(warmup='linear', warmup_iters=1000)
optimizer = dict(lr=0.025)
optimizer = dict(type='SGD', lr=0.025, momentum=0.9, weight_decay=0.0005)
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
train_dataloader = dict(batch_size=4, num_workers=4)
val_dataloader = dict(batch_size=4, num_workers=4)
test_dataloader = val_dataloader

3
configs/bisenetv1/bisenetv1_r18-d32_lr5e-3_4x4_512x512_160k_coco-stuff164k.py
View File

@ -10,4 +10,5 @@ model = dict(
dict(num_classes=171),
])
lr_config = dict(warmup='linear', warmup_iters=1000)
optimizer = dict(lr=0.005)
optimizer = dict(type='SGD', lr=0.005, momentum=0.9, weight_decay=0.0005)
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)

3
configs/bisenetv1/bisenetv1_r50-d32_4x4_1024x1024_160k_cityscapes.py
View File

@ -35,7 +35,8 @@ model = dict(
concat_input=False),
])
lr_config = dict(warmup='linear', warmup_iters=1000)
optimizer = dict(lr=0.05)
optimizer = dict(type='SGD', lr=0.05, momentum=0.9, weight_decay=0.0005)
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
train_dataloader = dict(batch_size=4, num_workers=4)
val_dataloader = dict(batch_size=4, num_workers=4)
test_dataloader = val_dataloader

3
configs/bisenetv1/bisenetv1_r50-d32_lr5e-3_4x4_512x512_160k_coco-stuff164k.py
View File

@ -15,4 +15,5 @@ model = dict(
dict(in_channels=512, channels=256, num_classes=171),
])
lr_config = dict(warmup='linear', warmup_iters=1000)
optimizer = dict(lr=0.005)
optimizer = dict(type='SGD', lr=0.005, momentum=0.9, weight_decay=0.0005)
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)

3
configs/bisenetv2/bisenetv2_fcn_4x4_1024x1024_160k_cityscapes.py
View File

@ -4,7 +4,8 @@ _base_ = [
'../_base_/default_runtime.py', '../_base_/schedules/schedule_160k.py'
]
lr_config = dict(warmup='linear', warmup_iters=1000)
optimizer = dict(lr=0.05)
optimizer = dict(type='SGD', lr=0.05, momentum=0.9, weight_decay=0.0005)
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
train_dataloader = dict(batch_size=4, num_workers=4)
val_dataloader = dict(batch_size=4, num_workers=4)
test_dataloader = val_dataloader

3
configs/bisenetv2/bisenetv2_fcn_4x8_1024x1024_160k_cityscapes.py
View File

@ -4,7 +4,8 @@ _base_ = [
'../_base_/default_runtime.py', '../_base_/schedules/schedule_160k.py'
]
lr_config = dict(warmup='linear', warmup_iters=1000)
optimizer = dict(lr=0.05)
optimizer = dict(type='SGD', lr=0.05, momentum=0.9, weight_decay=0.0005)
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
train_dataloader = dict(batch_size=8, num_workers=4)
val_dataloader = dict(batch_size=8, num_workers=4)
test_dataloader = val_dataloader

2
configs/bisenetv2/bisenetv2_fcn_fp16_4x4_1024x1024_160k_cityscapes.py
View File

@ -1,5 +1,5 @@
_base_ = './bisenetv2_fcn_4x4_1024x1024_160k_cityscapes.py'
# fp16 settings
optimizer_config = dict(type='Fp16OptimizerHook', loss_scale=512.)
default_hooks = dict(optimizer=dict(type='Fp16OptimizerHook', loss_scale=512.))
# fp16 placeholder
fp16 = dict()

63
configs/bisenetv2/bisenetv2_fcn_ohem_4x4_1024x1024_160k_cityscapes.py
View File

@ -3,9 +3,68 @@ _base_ = [
'../_base_/datasets/cityscapes_1024x1024.py',
'../_base_/default_runtime.py', '../_base_/schedules/schedule_160k.py'
]
sampler = dict(type='OHEMPixelSampler', thresh=0.7, min_kept=10000)
norm_cfg = dict(type='SyncBN', requires_grad=True)
models = dict(
decode_head=dict(
sampler=dict(type='OHEMPixelSampler', thresh=0.7, min_kept=10000)),
auxiliary_head=[
dict(
type='FCNHead',
in_channels=16,
channels=16,
num_convs=2,
num_classes=19,
in_index=1,
norm_cfg=norm_cfg,
concat_input=False,
align_corners=False,
sampler=dict(type='OHEMPixelSampler', thresh=0.7, min_kept=10000),
loss_decode=dict(
type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
dict(
type='FCNHead',
in_channels=32,
channels=64,
num_convs=2,
num_classes=19,
in_index=2,
norm_cfg=norm_cfg,
concat_input=False,
align_corners=False,
sampler=dict(type='OHEMPixelSampler', thresh=0.7, min_kept=10000),
loss_decode=dict(
type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
dict(
type='FCNHead',
in_channels=64,
channels=256,
num_convs=2,
num_classes=19,
in_index=3,
norm_cfg=norm_cfg,
concat_input=False,
align_corners=False,
sampler=dict(type='OHEMPixelSampler', thresh=0.7, min_kept=10000),
loss_decode=dict(
type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
dict(
type='FCNHead',
in_channels=128,
channels=1024,
num_convs=2,
num_classes=19,
in_index=4,
norm_cfg=norm_cfg,
concat_input=False,
align_corners=False,
sampler=dict(type='OHEMPixelSampler', thresh=0.7, min_kept=10000),
loss_decode=dict(
type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
],
)
lr_config = dict(warmup='linear', warmup_iters=1000)
optimizer = dict(lr=0.05)
optimizer = dict(type='SGD', lr=0.05, momentum=0.9, weight_decay=0.0005)
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
train_dataloader = dict(batch_size=4, num_workers=4)
val_dataloader = dict(batch_size=4, num_workers=4)
test_dataloader = val_dataloader

8
configs/cgnet/cgnet_512x1024_60k_cityscapes.py
View File

@ -2,12 +2,16 @@ _base_ = ['../_base_/models/cgnet.py', '../_base_/default_runtime.py']
# optimizer
optimizer = dict(type='Adam', lr=0.001, eps=1e-08, weight_decay=0.0005)
optimizer_config = dict()
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
# learning policy
lr_config = dict(policy='poly', power=0.9, min_lr=1e-4, by_epoch=False)
# runtime settings
total_iters = 60000
checkpoint_config = dict(by_epoch=False, interval=4000)
train_cfg = dict(
type='IterBasedTrainLoop', max_iters=total_iters, val_interval=4000)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')
default_hooks = dict(checkpoint=dict(by_epoch=False, interval=4000))
evaluation = dict(interval=4000, metric='mIoU')
# dataset settings

8
configs/cgnet/cgnet_680x680_60k_cityscapes.py
View File

@ -5,12 +5,16 @@ _base_ = [
# optimizer
optimizer = dict(type='Adam', lr=0.001, eps=1e-08, weight_decay=0.0005)
optimizer_config = dict()
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
# learning policy
lr_config = dict(policy='poly', power=0.9, min_lr=1e-4, by_epoch=False)
# runtime settings
total_iters = 60000
checkpoint_config = dict(by_epoch=False, interval=4000)
train_cfg = dict(
type='IterBasedTrainLoop', max_iters=total_iters, val_interval=4000)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')
default_hooks = dict(checkpoint=dict(by_epoch=False, interval=4000))
evaluation = dict(interval=4000, metric='mIoU')
img_norm_cfg = dict(

13
configs/convnext/upernet_convnext_base_fp16_512x512_160k_ade20k.py
View File

@ -10,17 +10,21 @@ model = dict(
)
optimizer = dict(
constructor='LearningRateDecayOptimizerConstructor',
_delete_=True,
type='AdamW',
lr=0.0001,
betas=(0.9, 0.999),
weight_decay=0.05,
weight_decay=0.05)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg={
'decay_rate': 0.9,
'decay_type': 'stage_wise',
'num_layers': 12
})
},
constructor='LearningRateDecayOptimizerConstructor')
lr_config = dict(
_delete_=True,
@ -37,6 +41,7 @@ train_dataloader = dict(batch_size=2)
val_dataloader = dict(batch_size=2)
test_dataloader = val_dataloader
# fp16 settings
optimizer_config = dict(type='Fp16OptimizerHook', loss_scale='dynamic')
default_hooks = dict(
optimizer=dict(type='Fp16OptimizerHook', loss_scale='dynamic'))
# fp16 placeholder
fp16 = dict()

13
configs/convnext/upernet_convnext_base_fp16_640x640_160k_ade20k.py
View File

@ -25,17 +25,21 @@ model = dict(
)
optimizer = dict(
constructor='LearningRateDecayOptimizerConstructor',
_delete_=True,
type='AdamW',
lr=0.0001,
betas=(0.9, 0.999),
weight_decay=0.05,
weight_decay=0.05)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg={
'decay_rate': 0.9,
'decay_type': 'stage_wise',
'num_layers': 12
})
},
constructor='LearningRateDecayOptimizerConstructor')
lr_config = dict(
_delete_=True,
@ -52,6 +56,7 @@ train_dataloader = dict(batch_size=2)
val_dataloader = dict(batch_size=2)
test_dataloader = val_dataloader
# fp16 settings
optimizer_config = dict(type='Fp16OptimizerHook', loss_scale='dynamic')
default_hooks = dict(
optimizer=dict(type='Fp16OptimizerHook', loss_scale='dynamic'))
# fp16 placeholder
fp16 = dict()

13
configs/convnext/upernet_convnext_large_fp16_640x640_160k_ade20k.py
View File

@ -25,17 +25,21 @@ model = dict(
)
optimizer = dict(
constructor='LearningRateDecayOptimizerConstructor',
_delete_=True,
type='AdamW',
lr=0.0001,
betas=(0.9, 0.999),
weight_decay=0.05,
weight_decay=0.05)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg={
'decay_rate': 0.9,
'decay_type': 'stage_wise',
'num_layers': 12
})
},
constructor='LearningRateDecayOptimizerConstructor')
lr_config = dict(
_delete_=True,
@ -52,6 +56,7 @@ train_dataloader = dict(batch_size=2)
val_dataloader = dict(batch_size=2)
test_dataloader = val_dataloader
# fp16 settings
optimizer_config = dict(type='Fp16OptimizerHook', loss_scale='dynamic')
default_hooks = dict(
optimizer=dict(type='Fp16OptimizerHook', loss_scale='dynamic'))
# fp16 placeholder
fp16 = dict()

13
configs/convnext/upernet_convnext_small_fp16_512x512_160k_ade20k.py
View File

@ -24,17 +24,21 @@ model = dict(
)
optimizer = dict(
constructor='LearningRateDecayOptimizerConstructor',
_delete_=True,
type='AdamW',
lr=0.0001,
betas=(0.9, 0.999),
weight_decay=0.05,
weight_decay=0.05)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg={
'decay_rate': 0.9,
'decay_type': 'stage_wise',
'num_layers': 12
})
},
constructor='LearningRateDecayOptimizerConstructor')
lr_config = dict(
_delete_=True,
@ -51,6 +55,7 @@ train_dataloader = dict(batch_size=2)
val_dataloader = dict(batch_size=2)
test_dataloader = val_dataloader
# fp16 settings
optimizer_config = dict(type='Fp16OptimizerHook', loss_scale='dynamic')
default_hooks = dict(
optimizer=dict(type='Fp16OptimizerHook', loss_scale='dynamic'))
# fp16 placeholder
fp16 = dict()

13
configs/convnext/upernet_convnext_tiny_fp16_512x512_160k_ade20k.py
View File

@ -24,17 +24,21 @@ model = dict(
)
optimizer = dict(
constructor='LearningRateDecayOptimizerConstructor',
_delete_=True,
type='AdamW',
lr=0.0001,
betas=(0.9, 0.999),
weight_decay=0.05,
weight_decay=0.05)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg={
'decay_rate': 0.9,
'decay_type': 'stage_wise',
'num_layers': 6
})
},
constructor='LearningRateDecayOptimizerConstructor')
lr_config = dict(
_delete_=True,
@ -51,6 +55,7 @@ train_dataloader = dict(batch_size=2)
val_dataloader = dict(batch_size=2)
test_dataloader = val_dataloader
# fp16 settings
optimizer_config = dict(type='Fp16OptimizerHook', loss_scale='dynamic')
default_hooks = dict(
optimizer=dict(type='Fp16OptimizerHook', loss_scale='dynamic'))
# fp16 placeholder
fp16 = dict()

13
configs/convnext/upernet_convnext_xlarge_fp16_640x640_160k_ade20k.py
View File

@ -25,17 +25,21 @@ model = dict(
)
optimizer = dict(
constructor='LearningRateDecayOptimizerConstructor',
_delete_=True,
type='AdamW',
lr=0.00008,
betas=(0.9, 0.999),
weight_decay=0.05,
weight_decay=0.05)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg={
'decay_rate': 0.9,
'decay_type': 'stage_wise',
'num_layers': 12
})
},
constructor='LearningRateDecayOptimizerConstructor')
lr_config = dict(
_delete_=True,
@ -52,6 +56,7 @@ train_dataloader = dict(batch_size=2)
val_dataloader = dict(batch_size=2)
test_dataloader = val_dataloader
# fp16 settings
optimizer_config = dict(type='Fp16OptimizerHook', loss_scale='dynamic')
default_hooks = dict(
optimizer=dict(type='Fp16OptimizerHook', loss_scale='dynamic'))
# fp16 placeholder
fp16 = dict()

2
configs/deeplabv3/deeplabv3_r101-d8_fp16_512x1024_80k_cityscapes.py
View File

@ -1,5 +1,5 @@
_base_ = './deeplabv3_r101-d8_512x1024_80k_cityscapes.py'
# fp16 settings
optimizer_config = dict(type='Fp16OptimizerHook', loss_scale=512.)
default_hooks = dict(optimizer=dict(type='Fp16OptimizerHook', loss_scale=512.))
# fp16 placeholder
fp16 = dict()

2
configs/deeplabv3plus/deeplabv3plus_r101-d8_fp16_512x1024_80k_cityscapes.py
View File

@ -1,5 +1,5 @@
_base_ = './deeplabv3plus_r101-d8_512x1024_80k_cityscapes.py'
# fp16 settings
optimizer_config = dict(type='Fp16OptimizerHook', loss_scale=512.)
default_hooks = dict(optimizer=dict(type='Fp16OptimizerHook', loss_scale=512.))
# fp16 placeholder
fp16 = dict()

6
configs/dpt/dpt_vit-b16_512x512_160k_ade20k.py
View File

@ -10,7 +10,11 @@ optimizer = dict(
type='AdamW',
lr=0.00006,
betas=(0.9, 0.999),
weight_decay=0.01,
weight_decay=0.01)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg=dict(
custom_keys={
'pos_embed': dict(decay_mult=0.),

2
configs/fcn/fcn_r101-d8_fp16_512x1024_80k_cityscapes.py
View File

@ -1,5 +1,5 @@
_base_ = './fcn_r101-d8_512x1024_80k_cityscapes.py'
# fp16 settings
optimizer_config = dict(type='Fp16OptimizerHook', loss_scale=512.)
default_hooks = dict(optimizer=dict(type='Fp16OptimizerHook', loss_scale=512.))
# fp16 placeholder
fp16 = dict()

7
configs/knet/knet_s3_deeplabv3_r50-d8_8x2_512x512_adamw_80k_ade20k.py
View File

@ -79,7 +79,12 @@ model = dict(
# optimizer
optimizer = dict(_delete_=True, type='AdamW', lr=0.0001, weight_decay=0.0005)
optimizer_config = dict(grad_clip=dict(max_norm=1, norm_type=2))
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
default_hooks = dict(
optimizer=dict(
_delete_=True,
type='OptimizerHook',
grad_clip=dict(max_norm=1, norm_type=2)))
# learning policy
lr_config = dict(
_delete_=True,

7
configs/knet/knet_s3_fcn_r50-d8_8x2_512x512_adamw_80k_ade20k.py
View File

@ -79,7 +79,12 @@ model = dict(
test_cfg=dict(mode='whole'))
# optimizer
optimizer = dict(_delete_=True, type='AdamW', lr=0.0001, weight_decay=0.0005)
optimizer_config = dict(grad_clip=dict(max_norm=1, norm_type=2))
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
default_hooks = dict(
optimizer=dict(
_delete_=True,
type='OptimizerHook',
grad_clip=dict(max_norm=1, norm_type=2)))
# learning policy
lr_config = dict(
_delete_=True,

7
configs/knet/knet_s3_pspnet_r50-d8_8x2_512x512_adamw_80k_ade20k.py
View File

@ -78,7 +78,12 @@ model = dict(
test_cfg=dict(mode='whole'))
# optimizer
optimizer = dict(_delete_=True, type='AdamW', lr=0.0001, weight_decay=0.0005)
optimizer_config = dict(grad_clip=dict(max_norm=1, norm_type=2))
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
default_hooks = dict(
optimizer=dict(
_delete_=True,
type='OptimizerHook',
grad_clip=dict(max_norm=1, norm_type=2)))
# learning policy
lr_config = dict(
_delete_=True,

7
configs/knet/knet_s3_upernet_r50-d8_8x2_512x512_adamw_80k_ade20k.py
View File

@ -79,7 +79,12 @@ model = dict(
test_cfg=dict(mode='whole'))
# optimizer
optimizer = dict(_delete_=True, type='AdamW', lr=0.0001, weight_decay=0.0005)
optimizer_config = dict(grad_clip=dict(max_norm=1, norm_type=2))
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
default_hooks = dict(
optimizer=dict(
_delete_=True,
type='OptimizerHook',
grad_clip=dict(max_norm=1, norm_type=2)))
# learning policy
lr_config = dict(
_delete_=True,

13
configs/knet/knet_s3_upernet_swin-t_8x2_512x512_adamw_80k_ade20k.py
View File

@ -36,14 +36,23 @@ optimizer = dict(
type='AdamW',
lr=0.00006,
betas=(0.9, 0.999),
weight_decay=0.0005,
weight_decay=0.0005)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg=dict(
custom_keys={
'absolute_pos_embed': dict(decay_mult=0.),
'relative_position_bias_table': dict(decay_mult=0.),
'norm': dict(decay_mult=0.)
}))
optimizer_config = dict(grad_clip=dict(max_norm=1, norm_type=2))
default_hooks = dict(
optimizer=dict(
_delete_=True,
type='OptimizerHook',
grad_clip=dict(max_norm=1, norm_type=2)))
# learning policy
lr_config = dict(
_delete_=True,

10
configs/mae/upernet_mae-base_fp16_8x2_512x512_160k_ade20k.py
View File

@ -27,9 +27,13 @@ optimizer = dict(
type='AdamW',
lr=1e-4,
betas=(0.9, 0.999),
weight_decay=0.05,
constructor='LayerDecayOptimizerConstructor',
paramwise_cfg=dict(num_layers=12, layer_decay_rate=0.65))
weight_decay=0.05)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg=dict(num_layers=12, layer_decay_rate=0.65),
constructor='LayerDecayOptimizerConstructor')
lr_config = dict(
_delete_=True,

3
configs/ocrnet/ocrnet_r101-d8_512x1024_40k_b16_cityscapes.py
View File

@ -3,5 +3,6 @@ _base_ = [
'../_base_/default_runtime.py', '../_base_/schedules/schedule_40k.py'
]
model = dict(pretrained='open-mmlab://resnet101_v1c', backbone=dict(depth=101))
optimizer = dict(lr=0.02)
optimizer = dict(type='SGD', lr=0.02, momentum=0.9, weight_decay=0.0005)
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
lr_config = dict(min_lr=2e-4)

3
configs/ocrnet/ocrnet_r101-d8_512x1024_80k_b16_cityscapes.py
View File

@ -3,5 +3,6 @@ _base_ = [
'../_base_/default_runtime.py', '../_base_/schedules/schedule_80k.py'
]
model = dict(pretrained='open-mmlab://resnet101_v1c', backbone=dict(depth=101))
optimizer = dict(lr=0.02)
optimizer = dict(type='SGD', lr=0.02, momentum=0.9, weight_decay=0.0005)
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
lr_config = dict(min_lr=2e-4)

2
configs/pspnet/pspnet_r101-d8_fp16_512x1024_80k_cityscapes.py
View File

@ -1,5 +1,5 @@
_base_ = './pspnet_r101-d8_512x1024_80k_cityscapes.py'
# fp16 settings
optimizer_config = dict(type='Fp16OptimizerHook', loss_scale=512.)
default_hooks = dict(optimizer=dict(type='Fp16OptimizerHook', loss_scale=512.))
# fp16 placeholder
fp16 = dict()

7
configs/pspnet/pspnet_r50-d32_rsb-pretrain_512x1024_adamw_80k_cityscapes.py
View File

@ -13,7 +13,12 @@ model = dict(
strides=(1, 2, 2, 2)))
optimizer = dict(_delete_=True, type='AdamW', lr=0.0005, weight_decay=0.05)
optimizer_config = dict(grad_clip=dict(max_norm=1, norm_type=2))
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
default_hooks = dict(
optimizer=dict(
_delete_=True,
type='OptimizerHook',
grad_clip=dict(max_norm=1, norm_type=2)))
# learning policy
lr_config = dict(
_delete_=True,

7
configs/pspnet/pspnet_r50-d8_rsb-pretrain_512x1024_adamw_80k_cityscapes.py
View File

@ -11,7 +11,12 @@ model = dict(
type='Pretrained', prefix='backbone.', checkpoint=checkpoint)))
optimizer = dict(_delete_=True, type='AdamW', lr=0.0005, weight_decay=0.05)
optimizer_config = dict(grad_clip=dict(max_norm=1, norm_type=2))
optim_wrapper = dict(type='OptimWrapper', optimizer=optimizer)
default_hooks = dict(
optimizer=dict(
_delete_=True,
type='OptimizerHook',
grad_clip=dict(max_norm=1, norm_type=2)))
# learning policy
lr_config = dict(
_delete_=True,

6
configs/segformer/segformer_mit-b0_512x512_160k_ade20k.py
View File

@ -12,7 +12,11 @@ optimizer = dict(
type='AdamW',
lr=0.00006,
betas=(0.9, 0.999),
weight_decay=0.01,
weight_decay=0.01)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg=dict(
custom_keys={
'pos_block': dict(decay_mult=0.),

7
configs/segformer/segformer_mit-b0_8x1_1024x1024_160k_cityscapes.py
View File

@ -15,14 +15,17 @@ optimizer = dict(
type='AdamW',
lr=0.00006,
betas=(0.9, 0.999),
weight_decay=0.01,
weight_decay=0.01)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg=dict(
custom_keys={
'pos_block': dict(decay_mult=0.),
'norm': dict(decay_mult=0.),
'head': dict(lr_mult=10.)
}))
lr_config = dict(
_delete_=True,
policy='poly',

6
configs/swin/upernet_swin_tiny_patch4_window7_512x512_160k_ade20k_pretrain_224x224_1K.py
View File

@ -23,7 +23,11 @@ optimizer = dict(
type='AdamW',
lr=0.00006,
betas=(0.9, 0.999),
weight_decay=0.01,
weight_decay=0.01)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg=dict(
custom_keys={
'absolute_pos_embed': dict(decay_mult=0.),

6
configs/twins/twins_pcpvt-s_uperhead_8x4_512x512_160k_ade20k.py
View File

@ -9,7 +9,11 @@ optimizer = dict(
type='AdamW',
lr=0.00006,
betas=(0.9, 0.999),
weight_decay=0.01,
weight_decay=0.01)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg=dict(custom_keys={
'pos_block': dict(decay_mult=0.),
'norm': dict(decay_mult=0.)

6
configs/twins/twins_svt-s_uperhead_8x2_512x512_160k_ade20k.py
View File

@ -24,7 +24,11 @@ optimizer = dict(
type='AdamW',
lr=0.00006,
betas=(0.9, 0.999),
weight_decay=0.01,
weight_decay=0.01)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg=dict(custom_keys={
'pos_block': dict(decay_mult=0.),
'norm': dict(decay_mult=0.)

6
configs/vit/upernet_vit-b16_ln_mln_512x512_160k_ade20k.py
View File

@ -17,7 +17,11 @@ optimizer = dict(
type='AdamW',
lr=0.00006,
betas=(0.9, 0.999),
weight_decay=0.01,
weight_decay=0.01)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg=dict(
custom_keys={
'pos_embed': dict(decay_mult=0.),

6
configs/vit/upernet_vit-b16_mln_512x512_160k_ade20k.py
View File

@ -16,7 +16,11 @@ optimizer = dict(
type='AdamW',
lr=0.00006,
betas=(0.9, 0.999),
weight_decay=0.01,
weight_decay=0.01)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg=dict(
custom_keys={
'pos_embed': dict(decay_mult=0.),

6
configs/vit/upernet_vit-b16_mln_512x512_80k_ade20k.py
View File

@ -16,7 +16,11 @@ optimizer = dict(
type='AdamW',
lr=0.00006,
betas=(0.9, 0.999),
weight_decay=0.01,
weight_decay=0.01)
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
paramwise_cfg=dict(
custom_keys={
'pos_embed': dict(decay_mult=0.),