Cspnext

configs/cspnext/README.md

@@ -0,0 +1,53 @@
+# CSPNeXt ImageNet Pre-training
+
+In this folder, we provide the imagenet pre-training config of RTMDet's backbone CSPNeXt.
+
+## Requirements
+
+To train with these configs, please install [MMClassification 1.x](https://github.com/open-mmlab/mmclassification/tree/1.x) first.
+
+Install by MIM:
+
+```shell
+mim install mmcls>=1.0.0rc0
+```
+
+or install by pip:
+
+```shell
+pip install mmcls>=1.0.0rc0
+```
+
+## Prepare Dataset
+
+To pre-train on ImageNet, you need to prepare the dataset first. Please refer to the [guide](https://mmclassification.readthedocs.io/en/1.x/user_guides/dataset_prepare.html#imagenet).
+
+## How to Train
+
+You can use the classification config in the same way as the detection config.
+
+For single-GPU training, run:
+
+```shell
+python tools/train.py \
+    ${CONFIG_FILE} \
+    [optional arguments]
+```
+
+For multi-GPU training, run:
+
+```shell
+bash ./tools/dist_train.sh \
+    ${CONFIG_FILE} \
+    ${GPU_NUM} \
+    [optional arguments]
+```
+
+More details can be found in [user guides](https://mmdetection.readthedocs.io/en/3.x/user_guides/train.html).
+
+## Results and Models
+
+|    Model     | resolution | Params(M) | Flops(G) | Top-1 (%) | Top-5 (%) |                                                      Download                                                       |
+| :----------: | :--------: | :-------: | :------: | :-------: | :-------: | :-----------------------------------------------------------------------------------------------------------------: |
+| CSPNeXt-tiny |  224x224   |   2.73    |  0.339   |   69.44   |   89.45   | [model](https://download.openmmlab.com/mmdetection/v3.0/rtmdet/cspnext_rsb_pretrain/cspnext-tiny_imagenet_600e.pth) |
+|  CSPNeXt-s   |  224x224   |   4.89    |  0.664   |   74.41   |   92.23   |  [model](https://download.openmmlab.com/mmdetection/v3.0/rtmdet/cspnext_rsb_pretrain/cspnext-s_imagenet_600e.pth)   |

configs/cspnext/cpsnext-tiny-600e_in100.py

@@ -0,0 +1,107 @@
+_base_ = './cspnext-s_8xb256-rsb-a1-600e_in1k.py'
+work_dir = "./work_dir/cspnext-tiny-600e_in100"
+# act_cfg = dict(type='ReLU6', inplace=True)
+
+model = dict(
+    data_preprocessor=dict(
+        num_classes=100,  # ImageNet-100 이므로 100으로 설정
+    ),
+    backbone=dict(deepen_factor=0.167, widen_factor=0.375),
+    head=dict(in_channels=384,
+    num_classes=100))
+
+train_pipeline = [
+    dict(type='LoadImageFromFile'),
+    dict(type='RandomResizedCrop', scale=224),
+    dict(type='RandomFlip', prob=0.5, direction='horizontal'),
+    dict(type='PackInputs'),  # <--- [필수]
+]
+
+test_pipeline = [
+    dict(type='LoadImageFromFile'),
+    dict(type='ResizeEdge', scale=256, edge='short'),
+    dict(type='CenterCrop', crop_size=224),
+    dict(type='PackInputs'),  # <--- [필수]
+]
+
+train_dataloader = dict(
+    batch_size=384,
+    num_workers=12,
+    dataset=dict(
+        type='ImageNet',
+        data_root='/mnt/hdd/OpenDataLab___ImageNet-100/raw/MyImagenet',
+        ann_file='meta/train.txt',
+        data_prefix='train/',
+        pipeline=train_pipeline), # <--- [추가]
+    sampler=dict(type='RepeatAugSampler', shuffle=True)
+)
+
+val_dataloader = dict(
+    batch_size=384,
+    num_workers=12,
+    dataset=dict(
+        type='ImageNet',
+        data_root='/mnt/hdd/OpenDataLab___ImageNet-100/raw/MyImagenet',
+        ann_file='meta/val.txt',
+        data_prefix='val/',
+        pipeline=test_pipeline), # <--- [추가]
+    sampler=dict(type='DefaultSampler', shuffle=False)
+)
+val_cfg = dict() 
+
+# 2. 평가 지표 설정 (정확도 Top-1, Top-5 측정)
+val_evaluator = dict(type='Accuracy', topk=(1, 5))
+
+# test_dataloader = val_dataloader
+# test_evaluator = val_evaluator
+
+vis_backends = [
+    dict(type='LocalVisBackend'),
+    dict(
+        type='WandbVisBackend',
+        init_kwargs=dict(
+            project='csp', # WandB 프로젝트 이름
+            name='CSPNeXt-tiny-default',      # (선택) 실험 이름
+            # entity='my-username'            # (선택) 팀/유저 이름
+        )
+    )
+]
+
+visualizer = dict(
+    type='UniversalVisualizer',  # <--- [중요] mmpretrain 전용 비주얼라이저 사용
+    vis_backends=vis_backends
+)
+
+default_hooks = dict(
+    # 기존 설정 유지
+    timer=dict(type='IterTimerHook'),
+    logger=dict(type='LoggerHook', interval=50),
+    param_scheduler=dict(type='ParamSchedulerHook'),
+    sampler_seed=dict(type='DistSamplerSeedHook'),
+
+    # [수정 1] Segmentation 전용 훅 제거 -> 일반 VisualizationHook 사용
+    visualization=dict(type='VisualizationHook', enable=True),
+
+    # [수정 2] CheckpointHook의 저장 기준 변경
+    checkpoint=dict(
+        type='CheckpointHook',
+        interval=1,                # 1 Epoch마다 저장 시도
+        save_best='accuracy/top1', # [변경] Segmentation(mIoU) -> Classification(Top-1 Accuracy)
+        rule='greater',            # 정확도는 높을수록 좋으므로 greater
+        max_keep_ckpts=3,          # Best 모델 최대 3개 유지 (주석에 따라 1로 줄여도 됨)
+        save_last=True             # 가장 마지막 Epoch 모델 별도 저장
+    )
+)
+
+optim_wrapper = dict(
+    type='OptimWrapper', # 기본 OptimWrapper 사용
+    optimizer=dict(
+        type='Lamb',      # 여기에 반드시 type이 있어야 합니다.
+        lr=0.005,
+        weight_decay=0.02
+    ),
+    paramwise_cfg=dict(
+        bias_decay_mult=0.0,
+        norm_decay_mult=0.0
+    )
+)

configs/cspnext/cspnext-s_8xb256-rsb-a1-600e_in1k.py

@@ -0,0 +1,67 @@
+_base_ = [
+    'mmcls::_base_/datasets/imagenet_bs256_rsb_a12.py',
+    'mmcls::_base_/schedules/imagenet_bs2048_rsb.py',
+    'mmcls::_base_/default_runtime.py'
+]
+default_scope = 'mmpretrain'
+# custom_imports = dict(
+#     imports=['mmdet.models', 'mmyolo.models'], allow_failed_imports=False)
+
+model = dict(
+    type='ImageClassifier',
+    backbone=dict(
+        type='CSPNeXt',
+        arch='P5',
+        out_indices=(4, ),
+        expand_ratio=0.5,
+        deepen_factor=0.33,
+        widen_factor=0.5,
+        channel_attention=True,
+        norm_cfg=dict(type='BN'),
+        act_cfg=dict(type='SiLU')),
+    neck=dict(type='GlobalAveragePooling'),
+    head=dict(
+        type='LinearClsHead',
+        num_classes=1000,
+        in_channels=512,
+        loss=dict(
+            type='LabelSmoothLoss',
+            label_smooth_val=0.1,
+            mode='original',
+            loss_weight=1.0),
+        topk=(1, 5)),
+    train_cfg=dict(augments=[
+        dict(type='Mixup', alpha=0.2),
+        dict(type='CutMix', alpha=1.0)
+    ]))
+
+# dataset settings
+train_dataloader = dict(sampler=dict(type='RepeatAugSampler', shuffle=True))
+
+# schedule settings
+optim_wrapper = dict(
+    optimizer=dict(weight_decay=0.01),
+    paramwise_cfg=dict(bias_decay_mult=0., norm_decay_mult=0.),
+)
+
+param_scheduler = [
+    # warm up learning rate scheduler
+    dict(
+        type='LinearLR',
+        start_factor=0.0001,
+        by_epoch=True,
+        begin=0,
+        end=5,
+        # update by iter
+        convert_to_iter_based=True),
+    # main learning rate scheduler
+    dict(
+        type='CosineAnnealingLR',
+        T_max=595,
+        eta_min=1.0e-6,
+        by_epoch=True,
+        begin=5,
+        end=600)
+]
+
+train_cfg = dict(by_epoch=True, max_epochs=600)

configs/cspnext/cspnext-tiny_8xb256-rsb-a1-600e_in1k.py

@@ -0,0 +1,5 @@
+_base_ = './cspnext-s_8xb256-rsb-a1-600e_in1k.py'
+
+model = dict(
+    backbone=dict(deepen_factor=0.167, widen_factor=0.375),
+    head=dict(in_channels=384))

mmpretrain/models/backbones/__init__.py

@@ -58,6 +58,7 @@
 from .vit_eva02 import ViTEVA02
 from .vit_sam import ViTSAM
 from .xcit import XCiT
+from .cspnext import CSPNeXt
 
 __all__ = [
     'LeNet5',
@@ -126,4 +127,5 @@
     'HiViT',
     'SparseResNet',
     'SparseConvNeXt',
+    'CSPNeXt',
 ]