How to prune yolo models in mmyolo with mmrazor？

[nuonepeaceyy]

Checklist

• I have searched related issues but cannot get the expected help.
• I have read related documents and don't know what to do.

Describe the question you meet

I try to configure the relevant yolov5s config files according to the example in configs/pruning/mmdet/dcff, but in fact, I don't have a very clear idea about how to do this. When I try to get the config template of target_pruning_ratio by using python tools/pruning/get_channel_units.py configs/pruning/mmdet/dcff/dcff_yolov5_s_cspdarknet_1b32.py --choice, it always get FileNotFoundError: [Errno 2] No such file or directory: '/tmp/tmpnbjbrug0/tmpi2cqdj_l.py'

Post related information

1. The output of pip list | grep "mmcv\|mmrazor\|^torch" mmcv 2.0.0rc4 mmdet 3.0.0rc5 mmrazor 1.0.0rc2 /home/MMLAB/mmrazor mmyolo 0.4.0 torch 1.12.1 torchaudio 0.12.1 torchvision 0.13.1
2. Your config file if you modified it or created a new one.

_base_ = [
    './dcff_yolov5_s_cspdarknet.py',
    # 'mmdet::_base_/datasets/coco_detection.py',
    # 'mmdet::_base_/schedules/schedule_2x.py',
    'mmyolo::_base_/default_runtime.py'
]

model = dict(
    _scope_='mmrazor',
    type='DCFF',
    architecture=_base_.architecture,
    mutator_cfg=dict(
        type='DCFFChannelMutator',
        channel_unit_cfg=dict(
            type='DCFFChannelUnit', default_args=dict(choice_mode='ratio')),
        parse_cfg=dict(
            type='ChannelAnalyzer',
            demo_input=(1, 3, 224, 224),
            tracer_type='FxTracer')),
    # target_pruning_ratio=target_pruning_ratio,
    step_freq=1,
    linear_schedule=False)

model_wrapper = dict(
    type='mmcv.MMDistributedDataParallel', find_unused_parameters=True)

val_cfg = dict(_delete_=True, type='mmrazor.ItePruneValLoop')

3. the detail in dcff_yolov5_s_cspdarknet.py.

# parameters that often need to be modified
num_classes = 80
img_scale = (640, 640)  # width, height
deepen_factor = 0.33
widen_factor = 0.5
max_epochs = 300
save_epoch_intervals = 10
train_batch_size_per_gpu = 16
train_num_workers = 8
val_batch_size_per_gpu = 1
val_num_workers = 2

# persistent_workers must be False if num_workers is 0.
persistent_workers = True

# Base learning rate for optim_wrapper
base_lr = 0.01

# only on Val
batch_shapes_cfg = dict(
    type='BatchShapePolicy',
    batch_size=val_batch_size_per_gpu,
    img_size=img_scale[0],
    size_divisor=32,
    extra_pad_ratio=0.5)

anchors = [
    [(10, 13), (16, 30), (33, 23)],  # P3/8
    [(30, 61), (62, 45), (59, 119)],  # P4/16
    [(116, 90), (156, 198), (373, 326)]  # P5/32
]
strides = [8, 16, 32]
num_det_layers = 3

# single-scale training is recommended to
# be turned on, which can speed up training.
env_cfg = dict(cudnn_benchmark=True)

architecture = dict(
    _scope_='mmyolo',
    type='YOLODetector',
    data_preprocessor=dict(  # 数据预处理器
        type='mmdet.DetDataPreprocessor',
        mean=[0., 0., 0.],
        std=[255., 255., 255.],
        bgr_to_rgb=True),
    backbone=dict(
        type='YOLOv5CSPDarknet',  # 主干网络
        deepen_factor=deepen_factor,
        widen_factor=widen_factor,
        norm_cfg=dict(type='BN', momentum=0.03, eps=0.001),
        act_cfg=dict(type='SiLU', inplace=True)),
    neck=dict(
        type='YOLOv5PAFPN',
        deepen_factor=deepen_factor,
        widen_factor=widen_factor,
        in_channels=[256, 512, 1024],
        out_channels=[256, 512, 1024],
        num_csp_blocks=3,
        norm_cfg=dict(type='BN', momentum=0.03, eps=0.001),
        act_cfg=dict(type='SiLU', inplace=True)),
    bbox_head=dict(
        type='YOLOv5Head',
        head_module=dict(
            type='YOLOv5HeadModule',
            num_classes=num_classes,
            in_channels=[256, 512, 1024],
            widen_factor=widen_factor,
            featmap_strides=strides,
            num_base_priors=3),
        prior_generator=dict(
            type='mmdet.YOLOAnchorGenerator',
            base_sizes=anchors,
            strides=strides),
        # scaled based on number of detection layers
        loss_cls=dict(
            type='mmdet.CrossEntropyLoss',
            use_sigmoid=True,
            reduction='mean',
            loss_weight=0.5 * (num_classes / 80 * 3 / num_det_layers)),
        loss_bbox=dict(
            type='IoULoss',
            iou_mode='ciou',
            bbox_format='xywh',
            eps=1e-7,
            reduction='mean',
            loss_weight=0.05 * (3 / num_det_layers),
            return_iou=True),
        loss_obj=dict(
            type='mmdet.CrossEntropyLoss',
            use_sigmoid=True,
            reduction='mean',
            loss_weight=1.0 * ((img_scale[0] / 640)**2 * 3 / num_det_layers)),
        prior_match_thr=4.,
        obj_level_weights=[4., 1., 0.4]),
    test_cfg=dict(
        multi_label=True,
        nms_pre=30000,
        score_thr=0.001,
        nms=dict(type='nms', iou_threshold=0.65),
        max_per_img=300))

4. the dcff_yolov5_s_cspdarknet.py and dcff_yolov5_s_cspdarknet_1b32.py are in the same path.

[nuonepeaceyy]

If I replace the './dcff_yolov5_s_cspdarknet.py' with './dcff_faster_rcnn_resnet50_fpn.py' in

_base_ = [
    './dcff_yolov5_s_cspdarknet.py',
    # 'mmdet::_base_/datasets/coco_detection.py',
    # 'mmdet::_base_/schedules/schedule_2x.py',
    'mmyolo::_base_/default_runtime.py'
]

this problem no longer exists, I don't understand the reason, they all exist under the same path and I don't know if my config file is correct.

[LKJacky]

When mmengine(openmmlab) raise error when a config can't be found, it usually print the absolute path of the config. Can you check the absolute path of the config?

[nuonepeaceyy]

When mmengine(openmmlab) raise error when a config can't be found, it usually print the absolute path of the config. Can you check the absolute path of the config?

Thank you for your answer, after changing the base file with mmyolo::yolov5/yolov5_s-v61_syncbn_fast_8xb16-300e_coco.py, the problem is solved.