How to change the training policy from Iter to Epoch based on the existing training profile ？如何根据已有的训练配置文件的训练政策从Iter改为Epoch？

[Xie-Muxi-BK]

For example：

configs/convnext/convnext-tiny_upernet_8xb2-amp-160k_ade20k-512x512.py

_base_ = [
'../_base_/models/upernet_convnext.py', '../_base_/datasets/ade20k.py',
'../_base_/default_runtime.py', '../_base_/schedules/schedule_160k.py'
]
crop_size = (512, 512)
data_preprocessor = dict(size=crop_size)
checkpoint_file = 'https://download.openmmlab.com/mmclassification/v0/convnext/downstream/convnext-tiny_3rdparty_32xb128-noema_in1k_20220301-795e9634.pth'  # noqa
model = dict(
data_preprocessor=data_preprocessor,
backbone=dict(
type='mmcls.ConvNeXt',
arch='tiny',
out_indices=[0, 1, 2, 3],
drop_path_rate=0.4,
layer_scale_init_value=1.0,
gap_before_final_norm=False,
init_cfg=dict(
type='Pretrained', checkpoint=checkpoint_file,
prefix='backbone.')),
decode_head=dict(
in_channels=[96, 192, 384, 768],
num_classes=150,
),
auxiliary_head=dict(in_channels=384, num_classes=150),
test_cfg=dict(mode='slide', crop_size=crop_size, stride=(341, 341)),
)

optim_wrapper = dict( delete=True, type='AmpOptimWrapper', optimizer=dict( type='AdamW', lr=0.0001, betas=(0.9, 0.999), weight_decay=0.05), paramwise_cfg={ 'decay_rate': 0.9, 'decay_type': 'stage_wise', 'num_layers': 6 }, constructor='LearningRateDecayOptimizerConstructor', loss_scale='dynamic')

param_scheduler = [ dict( type='LinearLR', start_factor=1e-6, by_epoch=False, begin=0, end=1500), dict( type='PolyLR', power=1.0, begin=1500, end=160000, eta_min=0.0, by_epoch=False, ) ]

By default, models are trained on 8 GPUs with 2 images per GPU

train_dataloader = dict(batch_size=2) val_dataloader = dict(batch_size=1) test_dataloader = val_dataloader

在mmseg0.X的修改runner的参数，但是现在没有runner了需要怎么改？
Modify runner parameters in mmseg0.X, but now there is no runner, what should be changed?

try:
```python
# training schedule for 160k
train_cfg = dict(
    type='EpochBasedTrainLoop', max_epochs=3, val_interval=1)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')

I have tried many things, but none of them worked properly Training always stays in the first epoch and does not even stop automatically

This is my completely modified configuration file：

norm_cfg = dict(type='SyncBN', requires_grad=True)
custom_imports = dict(imports='mmcls.models', allow_failed_imports=False)
checkpoint_file = 'https://download.openmmlab.com/mmclassification/v0/convnext/downstream/convnext-base_3rdparty_32xb128-noema_in1k_20220301-2a0ee547.pth'  # noqa
crop_size = (512, 512)

data_preprocessor = dict(
    type='SegDataPreProcessor',
    mean=[123.675, 116.28, 103.53],
    std=[58.395, 57.12, 57.375],
    bgr_to_rgb=True,
    pad_val=0,
    seg_pad_val=255,
    size=crop_size
)

model = dict(
    type='EncoderDecoder',
    data_preprocessor=data_preprocessor,
    pretrained=None,
    backbone=dict(
        type='mmcls.ConvNeXt',
        arch='base',
        out_indices=[0, 1, 2, 3],
        drop_path_rate=0.4,
        layer_scale_init_value=1.0,
        gap_before_final_norm=False,
        init_cfg=dict(
            type='Pretrained', checkpoint=checkpoint_file,
            prefix='backbone.')),
    decode_head=dict(
        type='UPerHead',
        in_channels=[128, 256, 512, 1024],
        in_index=[0, 1, 2, 3],
        pool_scales=(1, 2, 3, 6),
        channels=512,
        dropout_ratio=0.1,
        num_classes=150,
        norm_cfg=norm_cfg,
        align_corners=False,
        loss_decode=dict(
            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
    auxiliary_head=dict(
        type='FCNHead',
        in_channels=512,
        in_index=2,
        channels=256,
        num_convs=1,
        concat_input=False,
        dropout_ratio=0.1,
        num_classes=150,
        norm_cfg=norm_cfg,
        align_corners=False,
        loss_decode=dict(
            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=0.4)),
    # model training and testing settings
    train_cfg=dict(),
    test_cfg=dict(mode='slide', crop_size=crop_size, stride=(341, 341))
)

# dataset settings
dataset_type = 'ADE20KDataset'
data_root = 'data/ade/ADEChallengeData2016'
crop_size = (512, 512)
train_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='LoadAnnotations', reduce_zero_label=True),
    dict(
        type='RandomResize',
        scale=(2048, 512),
        ratio_range=(0.5, 2.0),
        keep_ratio=True),
    dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
    dict(type='RandomFlip', prob=0.5),
    dict(type='PhotoMetricDistortion'),
    dict(type='PackSegInputs')
]
test_pipeline = [
    dict(type='LoadImageFromFile'),
    dict(type='Resize', scale=(2048, 512), keep_ratio=True),
    # add loading annotation after ``Resize`` because ground truth
    # does not need to do resize data transform
    dict(type='LoadAnnotations', reduce_zero_label=True),
    dict(type='PackSegInputs')
]
img_ratios = [0.5, 0.75, 1.0, 1.25, 1.5, 1.75]
tta_pipeline = [
    dict(type='LoadImageFromFile', backend_args=None),
    dict(
        type='TestTimeAug',
        transforms=[
            [
                dict(type='Resize', scale_factor=r, keep_ratio=True)
                for r in img_ratios
            ],
            [
                dict(type='RandomFlip', prob=0., direction='horizontal'),
                dict(type='RandomFlip', prob=1., direction='horizontal')
            ], [dict(type='LoadAnnotations')], [dict(type='PackSegInputs')]
        ])
]
train_dataloader = dict(
    batch_size=2,
    num_workers=4,
    persistent_workers=True,
    sampler=dict(type='InfiniteSampler', shuffle=True),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        data_prefix=dict(
            img_path='images/training', seg_map_path='annotations/training'),
        pipeline=train_pipeline))
val_dataloader = dict(
    batch_size=1,
    num_workers=4,
    persistent_workers=True,
    sampler=dict(type='DefaultSampler', shuffle=False),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        data_prefix=dict(
            img_path='images/validation',
            seg_map_path='annotations/validation'),
        pipeline=test_pipeline))
test_dataloader = val_dataloader

val_evaluator = dict(type='IoUMetric', iou_metrics=['mIoU'])
test_evaluator = val_evaluator

# runtime settings
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'),
)
vis_backends = [dict(type='LocalVisBackend')]
visualizer = dict(
    type='SegLocalVisualizer', vis_backends=vis_backends, name='visualizer')
log_processor = dict(by_epoch=True)
log_level = 'INFO'
load_from = None
resume = False

tta_model = dict(type='SegTTAModel')

# optimizer
optim_wrapper = dict(
    # _delete_=True,
    type='AmpOptimWrapper',
    optimizer=dict(
        type='AdamW', lr=0.0001, betas=(0.9, 0.999), weight_decay=0.05),
    paramwise_cfg={
        'decay_rate': 0.9,
        'decay_type': 'stage_wise',
        'num_layers': 12
    },
    constructor='LearningRateDecayOptimizerConstructor',
    loss_scale='dynamic',
    clip_grad=None
)

# learning policy
param_scheduler = [
    dict(
        type='LinearLR', 
        start_factor=1e-6, 
        by_epoch=False, 
        begin=0, 
        end=1500
    ),
    dict(
        type='PolyLR',
        power=1.0,
        begin=1500,
        end=160000,
        eta_min=0.0,
        by_epoch=False,
    )
]

# training schedule for 160k
train_cfg = dict(
    type='EpochBasedTrainLoop', max_epochs=3, val_interval=1)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')
# default_hooks = dict(
#     # timer=dict(type='EpochTimerHook'),
#     # logger=dict(type='LoggerHook', interval=1, log_metric_by_epoch=True),
#     # param_scheduler=dict(type='ParamSchedulerHook'),
#     # checkpoint=dict(type='CheckpointHook', by_epoch=True, interval=1),
#     # sampler_seed=dict(type='DistSamplerSeedHook'),
#     visualization=dict(type='SegVisualizationHook'))

[huanruizhang123]

Hello, have you solved this problem? I am studying this problem, but I just learned this. I would like to ask if there is any difference between using Epoch and Iter, or just to get used to my own training method

[nightrain01]

You may need to modify the data configuration. Use 'DefaultSampler' instead of 'InfiniteSampler'.

train_dataloader = dict(
    batch_size=2,
    num_workers=4,
    persistent_workers=True,
    # sampler=dict(type='InfiniteSampler', shuffle=True),
    sampler=dict(type='DefaultSampler', shuffle=True),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        data_prefix=dict(
            img_path='images/training', seg_map_path='annotations/training'),
        pipeline=train_pipeline))

[gaohaozhang]

不知道这么做是不是有瑕疵，你可以看一下我得设置，可以运行但是细节我没有仔细研究： norm_cfg = dict(type='SyncBN', requires_grad=True) model = dict( type='EncoderDecoder', backbone=dict( type='ResNet', depth=50, num_stages=4, out_indices=(0, 1, 2, 3), dilations=(1, 1, 2, 4), strides=(1, 2, 1, 1), norm_cfg=dict(type='SyncBN', requires_grad=True), norm_eval=False, style='pytorch', contract_dilation=True), decode_head=dict( type='PSPHead', in_channels=2048, in_index=3, channels=512, pool_scales=(1, 2, 3, 6), dropout_ratio=0.1, num_classes=3, norm_cfg=dict(type='SyncBN', 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='slide', crop_size=(480, 480), stride=(320, 320))) dataset_type = 'PascalContextDataset' data_root = 'D:\MMlab\mmsegmentation-0.20.2\data\VOCdevkit\VOC2007' img_norm_cfg = dict(mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True) img_scale=(480, 480) crop_size = (480, 480) train_pipeline = [ dict(type='LoadImageFromFile'), dict(type='LoadAnnotations'), dict(type='Resize', img_scale=(480, 480), ratio_range=(1.0, 1.0)), dict(type='RandomCrop', crop_size=(480, 480), cat_max_ratio=0.75), dict(type='RandomFlip', prob=0.0), 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=(480, 480), 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=(480, 480), flip=False, transforms=[ dict(type='Resize', img_scale=(480, 480), ratio_range=(1.0, 1.0)), dict(type='RandomFlip', prob=0.0), 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=3, workers_per_gpu=1, train=dict( type='PascalContextDataset', data_root='D:\MMlab\mmsegmentation-0.20.2\data\VOCdevkit\VOC2007', img_dir='JPEGImages', ann_dir='SegmentationClassPNG', split='D:\MMlab\mmsegmentation-0.20.2\data\VOCdevkit\VOC2007\ImageSets\Segmentation\train.txt', pipeline=[ dict(type='LoadImageFromFile'), dict(type='LoadAnnotations'), dict(type='Resize', img_scale=(480, 480), ratio_range=(1.0, 1.0)), dict(type='RandomCrop', crop_size=(480, 480), cat_max_ratio=0.75), dict(type='RandomFlip', prob=0.0), 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=(480, 480), pad_val=0, seg_pad_val=255), dict(type='DefaultFormatBundle'), dict(type='Collect', keys=['img', 'gt_semantic_seg']) ]), val=dict( type='PascalContextDataset', data_root='D:\MMlab\mmsegmentation-0.20.2\data\VOCdevkit\VOC2007', img_dir='JPEGImages', ann_dir='SegmentationClassPNG', split='D:\MMlab\mmsegmentation-0.20.2\data\VOCdevkit\VOC2007\ImageSets\Segmentation\val.txt', pipeline=[ dict(type='LoadImageFromFile'), dict(type='MultiScaleFlipAug', img_scale=(480,480), flip=False, transforms=[ dict(type='Resize', img_scale=(480, 480), ratio_range=(1.0, 1.0)), dict(type='RandomFlip',prob=0.0), 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='PascalContextDataset', data_root='D:\MMlab\mmsegmentation-0.20.2\data\VOCdevkit\VOC2007', img_dir='JPEGImages', ann_dir='SegmentationClassPNG', split='D:\MMlab\mmsegmentation-0.20.2\data\VOCdevkit\VOC2007\ImageSets\Segmentation\test.txt', pipeline=[ dict(type='LoadImageFromFile'), dict( type='MultiScaleFlipAug', img_scale=(480,480), flip=False, transforms=[ dict(type='Resize', img_scale=(480, 480), ratio_range=(1.0, 1.0)), dict(type='RandomFlip',prob=0.0), 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=10, hooks=[dict(type='TextLoggerHook', by_epoch=True)]) dist_params = dict(backend='nccl') log_level = 'INFO' load_from = 'D:\MMlab\mmsegmentation-0.20.2\tools\work_dirs\pspnet_r50-d8_480x480_40k_pascal_context\latest.pth' resume_from = None workflow = [('train', 1)] cudnn_benchmark = True optimizer = dict(type='SGD', lr=0.004, momentum=0.9, weight_decay=0.0001) optimizer_config = dict() lr_config = dict(policy='poly', power=0.9, min_lr=0.0001, by_epoch=True) runner = dict(type='EpochBasedRunner', max_epochs=400) checkpoint_config = dict(interval=10,by_epoch=True) evaluation = dict(interval=1, metric='mIoU', pre_eval=True) work_dir = './work_dirs\pspnet_r50-d8text' test_evaluator = dict(type='IoUMetric', iou_metrics=['mIoU'], output_dir='work_dirs/format_results') gpu_ids = range(0, 1)

[yangtaowillv]

您好，请问您解决了吗？能否分享一下成功的经验？我也是读了mmengine的文档发现并不太适用于mmsegmentation 1.x