基于PaddlePaddle的ConvNeXt复现

│imagenet
├──train
│  ├── n01440764
│  │   ├── n01440764_10026.JPEG
│  │   ├── n01440764_10027.JPEG
│  │   ├── ......│  ├── ......├──val
│  ├── n01440764
│  │   ├── ILSVRC2012_val_00000293.JPEG
│  │   ├── ILSVRC2012_val_00002138.JPEG
│  │   ├── ......│  ├── ......

%cd /home/aistudio/# !git clone https://github.com/flytocc/ConvNeXt-paddle.git!unzip ConvNeXt-paddle-main.zip

# 需要安装2.2及以上版本的Paddle，如果# 安装GPU版本的Paddlepip install paddlepaddle-gpu==2.2.0# 安装CPU版本的Paddlepip install paddlepaddle==2.2.0

%cd /home/aistudio/ConvNeXt-paddle-main
!pip install -r requirements.txt

class Block(nn.Layer):
    r""" ConvNeXt Block. There are two equivalent implementations:
    (1) DwConv -> LayerNorm (channels_first) -> 1x1 Conv -> GELU -> 1x1 Conv; all in (N, C, H, W)
    (2) DwConv -> Permute to (N, H, W, C); LayerNorm (channels_last) -> Linear -> GELU -> Linear; Permute back
    We use (2) as we find it slightly faster in PyTorch
    Args:
        dim (int): Number of input channels.
        drop_path (float): Stochastic depth rate. Default: 0.0
        layer_scale_init_value (float): Init value for Layer Scale. Default: 1e-6.
    """

    def __init__(self, dim, drop_path=0., layer_scale_init_value=1e-6):
        super().__init__()
        self.dwconv = nn.Conv2D(dim, dim, 7, padding=3,
                                groups=dim)  # depthwise conv
        self.norm = nn.LayerNorm(dim, epsilon=1e-6)        # pointwise/1x1 convs, implemented with linear layers
        self.pwconv1 = nn.Linear(dim, 4 * dim)
        self.act = nn.GELU()
        self.pwconv2 = nn.Linear(4 * dim, dim)        if layer_scale_init_value > 0:
            self.gamma = self.create_parameter(
                shape=[dim],
                default_initializer=Constant(value=layer_scale_init_value))        else:
            self.gamma = None
        self.drop_path = DropPath(
            drop_path) if drop_path > 0. else nn.Identity()    def forward(self, x):
        input = x
        x = self.dwconv(x)
        x = x.transpose([0, 2, 3, 1])  # (N, C, H, W) -> (N, H, W, C)
        x = self.norm(x)
        x = self.pwconv1(x)
        x = self.act(x)
        x = self.pwconv2(x)        if self.gamma is not None:
            x = self.gamma * x
        x = x.transpose([0, 3, 1, 2])  # (N, H, W, C) -> (N, C, H, W)

        x = input + self.drop_path(x)        return x

%cd /home/aistudio/ConvNeXt-paddle-main

%run predict.py \
    --model convnext_tiny \
    --infer_imgs ./demo/ILSVRC2012_val_00020010.JPEG \
    --resume /home/aistudio/work/checkpoint-best.pd

[{'class_ids': [178, 211, 85, 236, 246], 'scores': [0.8764159083366394, 0.0005395704065449536, 0.0005327172111719847, 0.000466014607809484, 0.0004493744927458465], 'file_name': '/home/aistudio/ConvNeXt-paddle-main/demo/ILSVRC2012_val_00020010.JPEG', 'label_names': ['Weimaraner', 'vizsla, Hungarian pointer', 'quail', 'Doberman, Doberman pinscher', 'Great Dane']}]

export CUDA_VISIBLE_DEVICES=0,1,2,3
python -m paddle.distributed.launch --gpus="0,1,2,3" \
    main.py \
    --model convnext_tiny --drop_path 0.1 \
    --batch_size 128 --lr 4e-3 --accum_iter 8 \
    --warmup_epochs 20 \
    --model_ema --model_ema_eval --dist_eval \
    --data_path /path/to/imagenet/ \
    --cls_label_path_train /path/to/train_list.txt \
    --cls_label_path_val /path/to/val_list.txt \
    --output_dir output/convnext_tiny

[11:46:22.948892] Epoch: [96]  [ 840/2502]  eta: 0:15:25  lr: 0.003310  loss: 3.6854 (3.5704)  time: 0.5759  data: 0.0005[11:46:33.860486] Epoch: [96]  [ 860/2502]  eta: 0:15:14  lr: 0.003310  loss: 3.6475 (3.5700)  time: 0.5454  data: 0.0005

python eval.py \
    --model convnext_tiny \
    --batch_size 128 \
    --data_path /path/to/imagenet/ \
    --cls_label_path_val /path/to/val_list.txt \
    --resume $TRAINED_MODEL

%run export_model.py \
    --model convnext_tiny \
    --output_dir ./output/ \
    --resume /home/aistudio/work/checkpoint-best.pd

%run infer.py \
    --model_file ./output/model.pdmodel \
    --params_file ./output/model.pdiparams \
    --input_file ./demo/ILSVRC2012_val_00020010.JPEG

[{'class_ids': [178, 211, 85, 236, 246], 'scores': [0.876124918460846, 0.0005408977158367634, 0.0005338680348359048, 0.0004670217458624393, 0.0004502409719862044], 'file_name': './demo/ILSVRC2012_val_00020010.JPEG', 'label_names': ['Weimaraner', 'vizsla, Hungarian pointer', 'quail', 'Doberman, Doberman pinscher', 'Great Dane']}]

git clone https://github.com/LDOUBLEV/AutoLog
cd AutoLog/
pip3 install -r requirements.txt
python3 setup.py bdist_wheel
pip3 install ./dist/auto_log-1.2.0-py3-none-any.whl

bash test_tipc/prepare.sh test_tipc/config/ConvNeXt/convnext_tiny.txt 'lite_train_lite_infer'bash test_tipc/test_train_inference_python.sh test_tipc/config/ConvNeXt/convnext_tiny.txt 'lite_train_lite_infer'

Run successfully with command - python3.7 main.py --model=convnext_tiny --data_path=./dataset/ILSVRC2012/ --cls_label_path_train=./dataset/ILSVRC2012/train_list.txt --cls_label_path_val=./dataset/ILSVRC2012/val_list.txt --dist_eval    --output_dir=./test_tipc/output/norm_train_gpus_0_autocast_null/convnext_tiny --epochs=2     --batch_size=8 !Run successfully with command - python3.7 eval.py --model=convnext_tiny --data_path=./dataset/ILSVRC2012/ --cls_label_path_val=./dataset/ILSVRC2012/val_list.txt --resume=./test_tipc/output/norm_train_gpus_0_autocast_null/convnext_tiny/checkpoint-latest.pd !Run successfully with command - python3.7 export_model.py --model=convnext_tiny --resume=./test_tipc/output/norm_train_gpus_0_autocast_null/convnext_tiny/checkpoint-latest.pd --output=./test_tipc/output/norm_train_gpus_0_autocast_null !Run successfully with command - python3.7 infer.py --use_gpu=True --use_tensorrt=False --precision=fp32 --model_file=./test_tipc/output/norm_train_gpus_0_autocast_null/model.pdmodel --batch_size=1 --input_file=./dataset/ILSVRC2012/val  --params_file=./test_tipc/output/norm_train_gpus_0_autocast_null/model.pdiparams > ./test_tipc/output/python_infer_gpu_usetrt_False_precision_fp32_batchsize_1.log 2>&1 !......