Python torch.nn.init.uniform() Examples

The following are 30 code examples of torch.nn.init.uniform(). You can vote up the ones you like or vote down the ones you don't like, and go to the original project or source file by following the links above each example. You may also want to check out all available functions/classes of the module torch.nn.init , or try the search function .
Example #1
Source File: config.py    From pcl.pytorch with MIT License 6 votes vote down vote up 
def assert_and_infer_cfg(make_immutable=True):
    """Call this function in your script after you have finished setting all cfg
    values that are necessary (e.g., merging a config from a file, merging
    command line config options, etc.). By default, this function will also
    mark the global cfg as immutable to prevent changing the global cfg settings
    during script execution (which can lead to hard to debug errors or code
    that's harder to understand than is necessary).
    """
    if __C.MODEL.LOAD_IMAGENET_PRETRAINED_WEIGHTS:
        assert __C.VGG.IMAGENET_PRETRAINED_WEIGHTS, \
            "Path to the weight file must not be empty to load imagenet pertrained resnets."
    if version.parse(torch.__version__) < version.parse('0.4.0'):
        __C.PYTORCH_VERSION_LESS_THAN_040 = True
        # create alias for PyTorch version less than 0.4.0
        init.uniform_ = init.uniform
        init.normal_ = init.normal
        init.constant_ = init.constant
        nn.GroupNorm = mynn.GroupNorm
    if make_immutable:
        cfg.immutable(True)
Example #2
Source File: layers.py    From ASER with MIT License 6 votes vote down vote up 
def __init__(self, input_size, activation=nn.Tanh(),
                 method="dot"):
        super(AttnScore, self).__init__()
        self.activation = activation
        self.input_size = input_size
        self.method = method
        if method == "general":
            self.linear = nn.Linear(input_size, input_size)
            init.uniform(self.linear.weight.data, -0.005, 0.005)
        elif method == "concat":
            self.linear_1 = nn.Linear(input_size*2, input_size)
            self.linear_2 = nn.Linear(input_size, 1)
            init.uniform(self.linear_1.weight.data, -0.005, 0.005)
            init.uniform(self.linear_2.weight.data, -0.005, 0.005)
        elif method == "tri_concat":
            self.linear = nn.Linear(input_size*3, 1)
            init.uniform(self.linear.weight.data, -0.005, 0.005)
Example #3
Source File: flow_models.py    From hrnet with MIT License 5 votes vote down vote up 
def __init__(self, args, batchNorm=False, div_flow = 20.):
        super(FlowNet2CS,self).__init__()
        self.batchNorm = batchNorm
        self.div_flow = div_flow
        self.rgb_max = args.rgb_max
        self.args = args

        self.channelnorm = ChannelNorm()

        # First Block (FlowNetC)
        self.flownetc = FlowNetC.FlowNetC(args, batchNorm=self.batchNorm)
        self.upsample1 = nn.Upsample(scale_factor=4, mode='bilinear')

        if args.fp16:
            self.resample1 = nn.Sequential(
                            tofp32(), 
                            Resample2d(),
                            tofp16()) 
        else:
            self.resample1 = Resample2d()

        # Block (FlowNetS1)
        self.flownets_1 = FlowNetS.FlowNetS(args, batchNorm=self.batchNorm)
        self.upsample2 = nn.Upsample(scale_factor=4, mode='bilinear')

        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                if m.bias is not None:
                    init.uniform(m.bias)
                init.xavier_uniform(m.weight)

            if isinstance(m, nn.ConvTranspose2d):
                if m.bias is not None:
                    init.uniform(m.bias)
                init.xavier_uniform(m.weight)
                # init_deconv_bilinear(m.weight)
Example #4
Source File: cnn_train.py    From Evolutionary-Autoencoders with MIT License 5 votes vote down vote up 
def weights_init_kaiming(m):
    classname = m.__class__.__name__
    if classname.find('Conv') != -1:
        init.kaiming_normal(m.weight.data, a=0, mode='fan_in')
    elif classname.find('Linear') != -1:
        init.kaiming_normal(m.weight.data, a=0, mode='fan_in')
    elif classname.find('BatchNorm2d') != -1:
        init.uniform(m.weight.data, 1.0, 0.02)
        init.constant(m.bias.data, 0.0)
Example #5
Source File: cnn_train.py    From cgp-cnn-PyTorch with MIT License 5 votes vote down vote up 
def weights_init_orthogonal(m):
    classname = m.__class__.__name__
    print(classname)
    if classname.find('Conv') != -1:
        init.orthogonal(m.weight.data, gain=1)
    elif classname.find('Linear') != -1:
        init.orthogonal(m.weight.data, gain=1)
    elif classname.find('BatchNorm2d') != -1:
        init.uniform(m.weight.data, 1.0, 0.02)
        init.constant(m.bias.data, 0.0)
Example #6
Source File: cnn_train.py    From Evolutionary-Autoencoders with MIT License 5 votes vote down vote up 
def weights_init_normal_(m):
    classname = m.__class__.__name__
    if classname.find('Conv') != -1:
        init.uniform(m.weight.data, 0.0, 0.02)
    elif classname.find('Linear') != -1:
        init.uniform(m.weight.data, 0.0, 0.02)
    elif classname.find('BatchNorm2d') != -1:
        init.uniform(m.weight.data, 1.0, 0.02)
        init.constant(m.bias.data, 0.0)
Example #7
Source File: cnn_train.py    From Evolutionary-Autoencoders with MIT License 5 votes vote down vote up 
def weights_init_normal(m):
    classname = m.__class__.__name__
    if classname.find('Conv2d') != -1:
        m.apply(weights_init_normal_)
    elif classname.find('Linear') != -1:
        init.uniform(m.weight.data, 0.0, 0.02)
    elif classname.find('BatchNorm2d') != -1:
        init.uniform(m.weight.data, 1.0, 0.02)
        init.constant(m.bias.data, 0.0)
Example #8
Source File: models.py    From video-to-pose3D with MIT License 5 votes vote down vote up 
def __init__(self, args, batchNorm=False, div_flow=20.):
        super(FlowNet2CS, self).__init__()
        self.batchNorm = batchNorm
        self.div_flow = div_flow
        self.rgb_max = args.rgb_max
        self.args = args

        self.channelnorm = ChannelNorm()

        # First Block (FlowNetC)
        self.flownetc = FlowNetC.FlowNetC(args, batchNorm=self.batchNorm)
        self.upsample1 = nn.Upsample(scale_factor=4, mode='bilinear')

        if args.fp16:
            self.resample1 = nn.Sequential(
                tofp32(),
                Resample2d(),
                tofp16())
        else:
            self.resample1 = Resample2d()

        # Block (FlowNetS1)
        self.flownets_1 = FlowNetS.FlowNetS(args, batchNorm=self.batchNorm)
        self.upsample2 = nn.Upsample(scale_factor=4, mode='bilinear')

        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                if m.bias is not None:
                    init.uniform(m.bias)
                init.xavier_uniform(m.weight)

            if isinstance(m, nn.ConvTranspose2d):
                if m.bias is not None:
                    init.uniform(m.bias)
                init.xavier_uniform(m.weight)
                # init_deconv_bilinear(m.weight)
Example #9
Source File: cnn_train.py    From Evolutionary-Autoencoders with MIT License 5 votes vote down vote up 
def weights_init_xavier(m):
    classname = m.__class__.__name__
    if classname.find('Conv') != -1:
        init.xavier_normal(m.weight.data, gain=1)
    elif classname.find('Linear') != -1:
        init.xavier_normal(m.weight.data, gain=1)
    elif classname.find('BatchNorm2d') != -1:
        init.uniform(m.weight.data, 1.0, 0.02)
        init.constant(m.bias.data, 0.0)
Example #10
Source File: cnn_train.py    From Evolutionary-Autoencoders with MIT License 5 votes vote down vote up 
def weights_init_xavier(m):
    classname = m.__class__.__name__
    if classname.find('Conv') != -1:
        init.xavier_normal(m.weight.data, gain=1)
    elif classname.find('Linear') != -1:
        init.xavier_normal(m.weight.data, gain=1)
    elif classname.find('BatchNorm2d') != -1:
        init.uniform(m.weight.data, 1.0, 0.02)
        init.constant(m.bias.data, 0.0)
Example #11
Source File: cnn_train.py    From Evolutionary-Autoencoders with MIT License 5 votes vote down vote up 
def weights_init_kaiming(m):
    classname = m.__class__.__name__
    if classname.find('Conv') != -1:
        init.kaiming_normal(m.weight.data, a=0, mode='fan_in')
    elif classname.find('Linear') != -1:
        init.kaiming_normal(m.weight.data, a=0, mode='fan_in')
    elif classname.find('BatchNorm2d') != -1:
        init.uniform(m.weight.data, 1.0, 0.02)
        init.constant(m.bias.data, 0.0)
Example #12
Source File: cnn_train.py    From Evolutionary-Autoencoders with MIT License 5 votes vote down vote up 
def weights_init_orthogonal(m):
    classname = m.__class__.__name__
    print(classname)
    if classname.find('Conv') != -1:
        init.orthogonal(m.weight.data, gain=1)
    elif classname.find('Linear') != -1:
        init.orthogonal(m.weight.data, gain=1)
    elif classname.find('BatchNorm2d') != -1:
        init.uniform(m.weight.data, 1.0, 0.02)
        init.constant(m.bias.data, 0.0)
Example #13
Source File: cnn_train.py    From cgp-cnn-PyTorch with MIT License 5 votes vote down vote up 
def weights_init_normal(m):
    classname = m.__class__.__name__
    if classname.find('Conv2d') != -1:
        m.apply(weights_init_normal_)
    elif classname.find('Linear') != -1:
        init.uniform(m.weight.data, 0.0, 0.02)
    elif classname.find('BatchNorm2d') != -1:
        init.uniform(m.weight.data, 1.0, 0.02)
        init.constant(m.bias.data, 0.0)
Example #14
Source File: cnn_train.py    From cgp-cnn-PyTorch with MIT License 5 votes vote down vote up 
def weights_init_normal_(m):
    classname = m.__class__.__name__
    if classname.find('Conv') != -1:
        init.uniform(m.weight.data, 0.0, 0.02)
    elif classname.find('Linear') != -1:
        init.uniform(m.weight.data, 0.0, 0.02)
    elif classname.find('BatchNorm2d') != -1:
        init.uniform(m.weight.data, 1.0, 0.02)
        init.constant(m.bias.data, 0.0)
Example #15
Source File: cnn_train.py    From cgp-cnn-PyTorch with MIT License 5 votes vote down vote up 
def weights_init_xavier(m):
    classname = m.__class__.__name__
    if classname.find('Conv') != -1:
        init.xavier_normal(m.weight.data, gain=1)
    elif classname.find('Linear') != -1:
        init.xavier_normal(m.weight.data, gain=1)
    elif classname.find('BatchNorm2d') != -1:
        init.uniform(m.weight.data, 1.0, 0.02)
        init.constant(m.bias.data, 0.0)
Example #16
Source File: cnn_train.py    From cgp-cnn-PyTorch with MIT License 5 votes vote down vote up 
def weights_init_kaiming(m):
    classname = m.__class__.__name__
    if classname.find('Conv2d') != -1:
        init.kaiming_normal(m.weight.data, a=0, mode='fan_in')
    elif classname.find('Linear') != -1:
        init.kaiming_normal(m.weight.data, a=0, mode='fan_in')
    elif classname.find('BatchNorm2d') != -1:
        init.uniform(m.weight.data, 1.0, 0.02)
        init.constant(m.bias.data, 0.0)
Example #17
Source File: cnn_train.py    From Evolutionary-Autoencoders with MIT License 5 votes vote down vote up 
def weights_init_normal_(m):
    classname = m.__class__.__name__
    if classname.find('Conv') != -1:
        init.uniform(m.weight.data, 0.0, 0.02)
    elif classname.find('Linear') != -1:
        init.uniform(m.weight.data, 0.0, 0.02)
    elif classname.find('BatchNorm2d') != -1:
        init.uniform(m.weight.data, 1.0, 0.02)
        init.constant(m.bias.data, 0.0)
Example #18
Source File: models.py    From TreeEnc with MIT License 5 votes vote down vote up 
def reset_parameters(self):
        if self.use_batchnorm:
            self.bn_mlp_input.reset_parameters()
            self.bn_mlp_output.reset_parameters()
        for i in range(self.num_layers):
            linear_layer = self.mlp[i][0]
            init.kaiming_normal(linear_layer.weight.data)
            init.constant(linear_layer.bias.data, val=0)
        init.uniform(self.clf_linear.weight.data, -0.005, 0.005)
        init.constant(self.clf_linear.bias.data, val=0)
Example #19
Source File: config.py    From DIoU-pytorch-detectron with GNU General Public License v3.0 5 votes vote down vote up 
def assert_and_infer_cfg(make_immutable=True):
    """Call this function in your script after you have finished setting all cfg
    values that are necessary (e.g., merging a config from a file, merging
    command line config options, etc.). By default, this function will also
    mark the global cfg as immutable to prevent changing the global cfg settings
    during script execution (which can lead to hard to debug errors or code
    that's harder to understand than is necessary).
    """
    if __C.MODEL.RPN_ONLY or __C.MODEL.FASTER_RCNN:
        __C.RPN.RPN_ON = True
    if __C.RPN.RPN_ON or __C.RETINANET.RETINANET_ON:
        __C.TEST.PRECOMPUTED_PROPOSALS = False
    if __C.MODEL.LOAD_IMAGENET_PRETRAINED_WEIGHTS:
        assert __C.RESNETS.IMAGENET_PRETRAINED_WEIGHTS, \
            "Path to the weight file must not be empty to load imagenet pertrained resnets."
    if set([__C.MRCNN.ROI_MASK_HEAD, __C.KRCNN.ROI_KEYPOINTS_HEAD]) & _SHARE_RES5_HEADS:
        __C.MODEL.SHARE_RES5 = True
    if version.parse(torch.__version__) < version.parse('0.4.0'):
        __C.PYTORCH_VERSION_LESS_THAN_040 = True
        # create alias for PyTorch version less than 0.4.0
        init.uniform_ = init.uniform
        init.normal_ = init.normal
        init.constant_ = init.constant
        nn.GroupNorm = mynn.GroupNorm
    if make_immutable:
        cfg.immutable(True)
Example #20
Source File: networks.py    From Bayesian-CycleGAN with MIT License 5 votes vote down vote up 
def weights_init_uniform(m):
    classname = m.__class__.__name__
    if classname.find('Conv') != -1:
        init.uniform(m.weight.data, -0.06, 0.06)
    elif classname.find('Conv') != -1:
        init.uniform(m.weight.data, -0.06, 0.06)
    elif classname.find('BatchNorm2d') != -1:
        init.uniform(m.weight.data, 0.04, 1.06)
        init.constant(m.bias.data, 0.0)
Example #21
Source File: config.py    From detectron-self-train with MIT License 5 votes vote down vote up 
def assert_and_infer_cfg(make_immutable=False):
    """Call this function in your script after you have finished setting all cfg
    values that are necessary (e.g., merging a config from a file, merging
    command line config options, etc.). By default, this function will also
    mark the global cfg as immutable to prevent changing the global cfg settings
    during script execution (which can lead to hard to debug errors or code
    that's harder to understand than is necessary).
    """
    if __C.MODEL.RPN_ONLY or __C.MODEL.FASTER_RCNN:
        __C.RPN.RPN_ON = True
    if __C.RPN.RPN_ON or __C.RETINANET.RETINANET_ON:
        __C.TEST.PRECOMPUTED_PROPOSALS = False
    if __C.MODEL.LOAD_IMAGENET_PRETRAINED_WEIGHTS:
        assert __C.RESNETS.IMAGENET_PRETRAINED_WEIGHTS, \
            "Path to the weight file must not be empty to load imagenet pertrained resnets."
    if set([__C.MRCNN.ROI_MASK_HEAD, __C.KRCNN.ROI_KEYPOINTS_HEAD]) & _SHARE_RES5_HEADS:
        __C.MODEL.SHARE_RES5 = True
    if version.parse(torch.__version__) < version.parse('0.4.0'):
        __C.PYTORCH_VERSION_LESS_THAN_040 = True
        # create alias for PyTorch version less than 0.4.0
        init.uniform_ = init.uniform
        init.normal_ = init.normal
        init.constant_ = init.constant
        nn.GroupNorm = mynn.GroupNorm
    if make_immutable:
        cfg.immutable(True)
Example #22
Source File: config.py    From Large-Scale-VRD.pytorch with MIT License 5 votes vote down vote up 
def assert_and_infer_cfg(make_immutable=True):
    """Call this function in your script after you have finished setting all cfg
    values that are necessary (e.g., merging a config from a file, merging
    command line config options, etc.). By default, this function will also
    mark the global cfg as immutable to prevent changing the global cfg settings
    during script execution (which can lead to hard to debug errors or code
    that's harder to understand than is necessary).
    """
    if __C.MODEL.RPN_ONLY or __C.MODEL.FASTER_RCNN:
        __C.RPN.RPN_ON = True
    if __C.RPN.RPN_ON or __C.RETINANET.RETINANET_ON:
        __C.TEST.PRECOMPUTED_PROPOSALS = False
    if __C.MODEL.LOAD_IMAGENET_PRETRAINED_WEIGHTS:
        assert __C.RESNETS.IMAGENET_PRETRAINED_WEIGHTS or __C.VGG16.IMAGENET_PRETRAINED_WEIGHTS, \
            "Path to the weight file must not be empty to load imagenet pertrained resnets."
    if set([__C.MRCNN.ROI_MASK_HEAD, __C.KRCNN.ROI_KEYPOINTS_HEAD]) & _SHARE_RES5_HEADS:
        __C.MODEL.SHARE_RES5 = True
    if version.parse(torch.__version__) < version.parse('0.4.0'):
        __C.PYTORCH_VERSION_LESS_THAN_040 = True
        # create alias for PyTorch version less than 0.4.0
        init.uniform_ = init.uniform
        init.normal_ = init.normal
        init.constant_ = init.constant
        nn.GroupNorm = mynn.GroupNorm
    if make_immutable:
        cfg.immutable(True)
Example #23
Source File: config.py    From PMFNet with MIT License 5 votes vote down vote up 
def assert_and_infer_cfg(make_immutable=True):
    """Call this function in your script after you have finished setting all cfg
    values that are necessary (e.g., merging a config from a file, merging
    command line config options, etc.). By default, this function will also
    mark the global cfg as immutable to prevent changing the global cfg settings
    during script execution (which can lead to hard to debug errors or code
    that's harder to understand than is necessary).
    """
    if __C.MODEL.RPN_ONLY or __C.MODEL.FASTER_RCNN:
        __C.RPN.RPN_ON = True
    if __C.RPN.RPN_ON or __C.RETINANET.RETINANET_ON:
        __C.TEST.PRECOMPUTED_PROPOSALS = False
    if __C.MODEL.LOAD_IMAGENET_PRETRAINED_WEIGHTS:
        assert __C.RESNETS.IMAGENET_PRETRAINED_WEIGHTS, \
            "Path to the weight file must not be empty to load imagenet pertrained resnets."
    if set([__C.MRCNN.ROI_MASK_HEAD, __C.KRCNN.ROI_KEYPOINTS_HEAD]) & _SHARE_RES5_HEADS:
        __C.MODEL.SHARE_RES5 = True
    if version.parse(torch.__version__) < version.parse('0.4.0'):
        __C.PYTORCH_VERSION_LESS_THAN_040 = True
        # create alias for PyTorch version less than 0.4.0
        init.uniform_ = init.uniform
        init.normal_ = init.normal
        init.constant_ = init.constant
        nn.GroupNorm = mynn.GroupNorm
    if make_immutable:
        cfg.immutable(True)
Example #24
Source File: model.py    From NoisyNet-A3C with MIT License 5 votes vote down vote up 
def reset_parameters(self):
    if hasattr(self, 'sigma_weight'):  # Only init after all params added (otherwise super().__init__() fails)
      init.uniform(self.weight, -math.sqrt(3 / self.in_features), math.sqrt(3 / self.in_features))
      init.uniform(self.bias, -math.sqrt(3 / self.in_features), math.sqrt(3 / self.in_features))
      init.constant(self.sigma_weight, self.sigma_init)
      init.constant(self.sigma_bias, self.sigma_init)
Example #25
Source File: config.py    From PANet with MIT License 5 votes vote down vote up 
def assert_and_infer_cfg(make_immutable=True):
    """Call this function in your script after you have finished setting all cfg
    values that are necessary (e.g., merging a config from a file, merging
    command line config options, etc.). By default, this function will also
    mark the global cfg as immutable to prevent changing the global cfg settings
    during script execution (which can lead to hard to debug errors or code
    that's harder to understand than is necessary).
    """
    if __C.MODEL.RPN_ONLY or __C.MODEL.FASTER_RCNN:
        __C.RPN.RPN_ON = True
    if __C.RPN.RPN_ON or __C.RETINANET.RETINANET_ON:
        __C.TEST.PRECOMPUTED_PROPOSALS = False
    if __C.MODEL.LOAD_IMAGENET_PRETRAINED_WEIGHTS:
        assert __C.RESNETS.IMAGENET_PRETRAINED_WEIGHTS, \
            "Path to the weight file must not be empty to load imagenet pertrained resnets."
    if set([__C.MRCNN.ROI_MASK_HEAD, __C.KRCNN.ROI_KEYPOINTS_HEAD]) & _SHARE_RES5_HEADS:
        __C.MODEL.SHARE_RES5 = True
    if version.parse(torch.__version__) < version.parse('0.4.0'):
        __C.PYTORCH_VERSION_LESS_THAN_040 = True
        # create alias for PyTorch version less than 0.4.0
        init.uniform_ = init.uniform
        init.normal_ = init.normal
        init.constant_ = init.constant
        nn.GroupNorm = mynn.GroupNorm
    if make_immutable:
        cfg.immutable(True)
Example #26
Source File: models.py    From RCRNet-Pytorch with MIT License 5 votes vote down vote up 
def __init__(self, args, batchNorm=False, div_flow = 20.):
        super(FlowNet2CS,self).__init__()
        self.batchNorm = batchNorm
        self.div_flow = div_flow
        self.rgb_max = args.rgb_max
        self.args = args

        self.channelnorm = ChannelNorm()

        # First Block (FlowNetC)
        self.flownetc = FlowNetC.FlowNetC(args, batchNorm=self.batchNorm)
        self.upsample1 = nn.Upsample(scale_factor=4, mode='bilinear')

        if args.fp16:
            self.resample1 = nn.Sequential(
                            tofp32(), 
                            Resample2d(),
                            tofp16()) 
        else:
            self.resample1 = Resample2d()

        # Block (FlowNetS1)
        self.flownets_1 = FlowNetS.FlowNetS(args, batchNorm=self.batchNorm)
        self.upsample2 = nn.Upsample(scale_factor=4, mode='bilinear')

        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                if m.bias is not None:
                    init.uniform(m.bias)
                init.xavier_uniform(m.weight)

            if isinstance(m, nn.ConvTranspose2d):
                if m.bias is not None:
                    init.uniform(m.bias)
                init.xavier_uniform(m.weight)
                # init_deconv_bilinear(m.weight)
Example #27
Source File: config.py    From Context-aware-ZSR with MIT License 5 votes vote down vote up 
def assert_and_infer_cfg(make_immutable=True):
    """Call this function in your script after you have finished setting all cfg
    values that are necessary (e.g., merging a config from a file, merging
    command line config options, etc.). By default, this function will also
    mark the global cfg as immutable to prevent changing the global cfg settings
    during script execution (which can lead to hard to debug errors or code
    that's harder to understand than is necessary).
    """
    if __C.MODEL.RPN_ONLY or __C.MODEL.FASTER_RCNN:
        __C.RPN.RPN_ON = True
    if __C.MODEL.TAGGING and __C.TEST.TAGGING:
        __C.TEST.USE_GT_PROPOSALS = True
    if __C.TEST.USE_GT_PROPOSALS:
        __C.MODEL.FASTER_RCNN = False # TODO: Currently this will not affect the training; It will potentially influence FPN if we do FPN
        __C.TEST.BBOX_REG = False
    if __C.MODEL.RELATION_COOCCUR:
        __C.MODEL.NUM_RELATIONS = 2
    if __C.RPN.RPN_ON or __C.RETINANET.RETINANET_ON or __C.TEST.USE_GT_PROPOSALS:
        __C.TEST.PRECOMPUTED_PROPOSALS = False
    if __C.MODEL.LOAD_IMAGENET_PRETRAINED_WEIGHTS:
        assert __C.RESNETS.IMAGENET_PRETRAINED_WEIGHTS or __C.TRAIN.IMAGENET_PRETRAINED_WEIGHTS, \
            "Path to the weight file must not be empty to load imagenet pertrained resnets."
    if set([__C.MRCNN.ROI_MASK_HEAD, __C.KRCNN.ROI_KEYPOINTS_HEAD]) & _SHARE_RES5_HEADS:
        __C.MODEL.SHARE_RES5 = True
    if version.parse(torch.__version__) < version.parse('0.4.0'):
        __C.PYTORCH_VERSION_LESS_THAN_040 = True
        # create alias for PyTorch version less than 0.4.0
        init.uniform_ = init.uniform
        init.normal_ = init.normal
        init.constant_ = init.constant
        nn.GroupNorm = mynn.GroupNorm
    if make_immutable:
        cfg.immutable(True)
Example #28
Source File: config.py    From Detectron.pytorch with MIT License 5 votes vote down vote up 
def assert_and_infer_cfg(make_immutable=True):
    """Call this function in your script after you have finished setting all cfg
    values that are necessary (e.g., merging a config from a file, merging
    command line config options, etc.). By default, this function will also
    mark the global cfg as immutable to prevent changing the global cfg settings
    during script execution (which can lead to hard to debug errors or code
    that's harder to understand than is necessary).
    """
    if __C.MODEL.RPN_ONLY or __C.MODEL.FASTER_RCNN:
        __C.RPN.RPN_ON = True
    if __C.RPN.RPN_ON or __C.RETINANET.RETINANET_ON:
        __C.TEST.PRECOMPUTED_PROPOSALS = False
    if __C.MODEL.LOAD_IMAGENET_PRETRAINED_WEIGHTS:
        assert __C.RESNETS.IMAGENET_PRETRAINED_WEIGHTS, \
            "Path to the weight file must not be empty to load imagenet pertrained resnets."
    if set([__C.MRCNN.ROI_MASK_HEAD, __C.KRCNN.ROI_KEYPOINTS_HEAD]) & _SHARE_RES5_HEADS:
        __C.MODEL.SHARE_RES5 = True
    if version.parse(torch.__version__) < version.parse('0.4.0'):
        __C.PYTORCH_VERSION_LESS_THAN_040 = True
        # create alias for PyTorch version less than 0.4.0
        init.uniform_ = init.uniform
        init.normal_ = init.normal
        init.constant_ = init.constant
        nn.GroupNorm = mynn.GroupNorm
    if make_immutable:
        cfg.immutable(True)
Example #29
Source File: weight_initialization.py    From aerial_mtl with BSD 3-Clause "New" or "Revised" License 5 votes vote down vote up 
def init_weights(net, init_type='normal'):
    print('initialization method [%s]' % init_type)
    if init_type == 'normal':
        net.apply(weights_init_normal)
    elif init_type == 'uniform':
        net.apply(weights_init_uniform)
    elif init_type == 'xavier':
        net.apply(weights_init_xavier)
    elif init_type == 'kaiming':
        net.apply(weights_init_kaiming)
    elif init_type == 'orthogonal':
        net.apply(weights_init_orthogonal)
    else:
        raise NotImplementedError('initialization method [%s] is not implemented' % init_type)
Example #30
Source File: weight_initialization.py    From aerial_mtl with BSD 3-Clause "New" or "Revised" License 5 votes vote down vote up 
def weights_init_orthogonal(m):
    classname = m.__class__.__name__
    print(classname)
    if classname.find('Conv') != -1:
        init.orthogonal(m.weight.data, gain=1)
    elif classname.find('Linear') != -1:
        init.orthogonal(m.weight.data, gain=1)
    elif classname.find('BatchNorm2d') != -1:
        init.uniform(m.weight.data, 1.0, 0.02)
        init.constant(m.bias.data, 0.0)