sadaco.apis.models package

Submodules

sadaco.apis.models.build_model module

sadaco.apis.models.build_model.build_model(configs)[source]

sadaco.apis.models.cbam module

class sadaco.apis.models.cbam.BasicConv(in_planes, out_planes, kernel_size, stride=1, padding=0, dilation=1, groups=1, relu=True, bn=True, bias=False)[source]

Bases: torch.nn.modules.module.Module

forward(x)[source]
training: bool
class sadaco.apis.models.cbam.Flatten[source]

Bases: torch.nn.modules.module.Module

forward(x)[source]
training: bool
class sadaco.apis.models.cbam.ChannelGate(gate_channels, reduction_ratio=16, pool_types=['avg', 'max'])[source]

Bases: torch.nn.modules.module.Module

forward(x)[source]
training: bool
sadaco.apis.models.cbam.logsumexp_2d(tensor)[source]
class sadaco.apis.models.cbam.ChannelPool[source]

Bases: torch.nn.modules.module.Module

forward(x)[source]
training: bool
class sadaco.apis.models.cbam.SpatialGate[source]

Bases: torch.nn.modules.module.Module

forward(x)[source]
training: bool
class sadaco.apis.models.cbam.CBAM(gate_channels, reduction_ratio=16, pool_types=['avg', 'max'], no_spatial=False)[source]

Bases: torch.nn.modules.module.Module

forward(x)[source]
training: bool

sadaco.apis.models.cnn_moe module

sadaco.apis.models.cnn_moe.cnn_moe(num_classes: int, num_experts: int = 10, cfg: Dict[str, List[Union[str, int, float]]] = {'conv1': [1, 64, 'AP', 0.1], 'conv2': [64, 128, 'AP', 0.15], 'conv3': [128, 256, None, 0.2], 'conv4': [256, 256, 'AP', 0.2], 'conv5': [256, 512, None, 0.25], 'conv6': [512, 512, None, None]}, **kwargs: Any) → sadaco.apis.models.cnn_moe.CNN_MoE[source]

Create CNN-MoE model from DCNN and Mixture of Experts.

Parameters

  • num_classes – Number of classes.

  • num_experts – Number of experts. (default: 10)

  • cfg – Block configuration. (default: CFG)

sadaco.apis.models.compile_trt module

sadaco.apis.models.compile_trt.compile(model: torch.nn.modules.module.Module, input_shape: tuple = 1, 3, 224, 224, device_id=0, batch_size=1, checkpoint=None, output_names=None)[source]

sadaco.apis.models.custom module

sadaco.apis.models.custom.custom_model(yml_path)[source]

sadaco.apis.models.resnet module

sadaco.apis.models.resnet.conv3x3(in_planes, out_planes, stride=1)[source]

3x3 convolution with padding

class sadaco.apis.models.resnet.BasicBlock(inplanes, planes, stride=1, downsample=None, use_cbam=False)[source]

Bases: torch.nn.modules.module.Module

expansion = 1
forward(x)[source]
training: bool
class sadaco.apis.models.resnet.Bottleneck(inplanes, planes, stride=1, downsample=None, use_cbam=False)[source]

Bases: torch.nn.modules.module.Module

expansion = 4
forward(x)[source]
training: bool
class sadaco.apis.models.resnet.ResNet(block, layers, network_type, num_classes, att_type, in_channel)[source]

Bases: torch.nn.modules.module.Module

training: bool
forward(x)[source]
sadaco.apis.models.resnet.ResidualNet(network_type, depth, num_classes, att_type, in_channel=3, pretrained=False)[source]

Module contents