lamhalobotnet50ts_256.a1h_in1k模型卡片

一个基于ResNet架构的Lambda+Halo+BoTNet图像分类模型。由Ross Wightman在ImageNet-1k上使用timm进行训练。

注意：该模型没有遵循任何特定论文的配置，它经过调整以获得合理的训练时间和降低自注意力块的频率。

配方详细信息：

• 基于 ResNet Strikes Back A1配方
• LAMB优化器
• 比论文A1配方具有更强的丢弃、随机深度和RandAugment
• 余弦学习率衰减策略与预热

该模型架构使用timm的灵活性 BYOBNet (Bring-Your-Own-Blocks Network) 实现。

BYOB（具有BYOANet注意力专用块）允许配置:

• 块/阶段布局
• 块类型交错
• 干细胞布局
• 输出步幅（膨胀）
• 激活层和规范层
• 通道和空间/自注意层

...并且还包括timm中许多其他架构共有的特点，包括:

• 随机深度
• 梯度检查点
• 逐层学习率衰减
• 阶段性特征提取

模型细节

• 模型类型：图像分类/特征主干
• 模型统计：
  • 参数（M）：22.6
  • GMACs：5.0
  • 激活数（M）：18.4
  • 图像大小：256 x 256
• 论文：
  • LambdaNetworks: 模拟无自注意的长程交互: https://arxiv.org/abs/2102.08602
  • 为参数高效的视觉骨干进行局部自注意的缩放： https://arxiv.org/abs/2103.12731
  • 用于视觉识别的瓶颈变换器： https://arxiv.org/abs/2101.11605
  • ResNet回击：在timm中改进的训练过程： https://arxiv.org/abs/2110.00476
• 数据集：ImageNet-1k

模型用途

图像分类

from urllib.request import urlopen
from PIL import Image
import timm

img = Image.open(urlopen(
    'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'
))

model = timm.create_model('lamhalobotnet50ts_256.a1h_in1k', pretrained=True)
model = model.eval()

# get model specific transforms (normalization, resize)
data_config = timm.data.resolve_model_data_config(model)
transforms = timm.data.create_transform(**data_config, is_training=False)

output = model(transforms(img).unsqueeze(0))  # unsqueeze single image into batch of 1

top5_probabilities, top5_class_indices = torch.topk(output.softmax(dim=1) * 100, k=5)

特征图提取

from urllib.request import urlopen
from PIL import Image
import timm

img = Image.open(urlopen(
    'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'
))

model = timm.create_model(
    'lamhalobotnet50ts_256.a1h_in1k',
    pretrained=True,
    features_only=True,
)
model = model.eval()

# get model specific transforms (normalization, resize)
data_config = timm.data.resolve_model_data_config(model)
transforms = timm.data.create_transform(**data_config, is_training=False)

output = model(transforms(img).unsqueeze(0))  # unsqueeze single image into batch of 1

for o in output:
    # print shape of each feature map in output
    # e.g.:
    #  torch.Size([1, 32, 128, 128])
    #  torch.Size([1, 256, 64, 64])
    #  torch.Size([1, 512, 32, 32])
    #  torch.Size([1, 1024, 16, 16])
    #  torch.Size([1, 2048, 8, 8])

    print(o.shape)

图像嵌入

from urllib.request import urlopen
from PIL import Image
import timm

img = Image.open(urlopen(
    'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'
))

model = timm.create_model(
    'lamhalobotnet50ts_256.a1h_in1k',
    pretrained=True,
    num_classes=0,  # remove classifier nn.Linear
)
model = model.eval()

# get model specific transforms (normalization, resize)
data_config = timm.data.resolve_model_data_config(model)
transforms = timm.data.create_transform(**data_config, is_training=False)

output = model(transforms(img).unsqueeze(0))  # output is (batch_size, num_features) shaped tensor

# or equivalently (without needing to set num_classes=0)

output = model.forward_features(transforms(img).unsqueeze(0))
# output is unpooled, a (1, 2048, 8, 8) shaped tensor

output = model.forward_head(output, pre_logits=True)
# output is a (1, num_features) shaped tensor

模型比较

在timm的 model results 中探索该模型的数据集和运行时指标。

引用

@misc{rw2019timm,
  author = {Ross Wightman},
  title = {PyTorch Image Models},
  year = {2019},
  publisher = {GitHub},
  journal = {GitHub repository},
  doi = {10.5281/zenodo.4414861},
  howpublished = {\url{https://github.com/huggingface/pytorch-image-models}}
}

@article{Bello2021LambdaNetworksML,
  title={LambdaNetworks: Modeling Long-Range Interactions Without Attention},
  author={Irwan Bello},
  journal={ArXiv},
  year={2021},
  volume={abs/2102.08602}
}

@article{Vaswani2021ScalingLS,
  title={Scaling Local Self-Attention for Parameter Efficient Visual Backbones},
  author={Ashish Vaswani and Prajit Ramachandran and A. Srinivas and Niki Parmar and Blake A. Hechtman and Jonathon Shlens},
  journal={2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2021},
  pages={12889-12899}
}

@article{Srinivas2021BottleneckTF,
  title={Bottleneck Transformers for Visual Recognition},
  author={A. Srinivas and Tsung-Yi Lin and Niki Parmar and Jonathon Shlens and P. Abbeel and Ashish Vaswani},
  journal={2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2021},
  pages={16514-16524}
}

@inproceedings{wightman2021resnet,
  title={ResNet strikes back: An improved training procedure in timm},
  author={Wightman, Ross and Touvron, Hugo and Jegou, Herve},
  booktitle={NeurIPS 2021 Workshop on ImageNet: Past, Present, and Future}
}