Python torch.nn.init.uniform_() Examples

def __init__(self, features, orthogonal_initialization=True, using_cache=False):
        """Constructor.

        Args:
            features: int, number of input features.
            orthogonal_initialization: bool, if True initialize weights to be a random
                orthogonal matrix.

        Raises:
            TypeError: if `features` is not a positive integer.
        """
        super().__init__(features, using_cache)

        if orthogonal_initialization:
            self._weight = nn.Parameter(utils.random_orthogonal(features))
        else:
            self._weight = nn.Parameter(torch.empty(features, features))
            stdv = 1.0 / np.sqrt(features)
            init.uniform_(self._weight, -stdv, stdv) 

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) 

def __init__(self, context_dim, state_dim, input_dim, bias=True):
        super().__init__()
        self._v_c = nn.Parameter(torch.Tensor(context_dim))
        self._v_s = nn.Parameter(torch.Tensor(state_dim))
        self._v_i = nn.Parameter(torch.Tensor(input_dim))
        init.uniform_(self._v_c, -INIT, INIT)
        init.uniform_(self._v_s, -INIT, INIT)
        init.uniform_(self._v_i, -INIT, INIT)
        if bias:
            self._b = nn.Parameter(torch.zeros(1))
        else:
            self.register_parameter(None, '_b') 

def reset_parameters(self):
        self.reset_running_stats()
        if self.affine:
            init.uniform_(self.weight)
            init.zeros_(self.bias) 

def __init__(self, emb_size, emb_dimension):
        super(SkipGramModel, self).__init__()
        self.emb_size = emb_size
        self.emb_dimension = emb_dimension
        self.u_embeddings = nn.Embedding(emb_size, emb_dimension, sparse=True)
        self.v_embeddings = nn.Embedding(emb_size, emb_dimension, sparse=True)

        initrange = 1.0 / self.emb_dimension
        init.uniform_(self.u_embeddings.weight.data, -initrange, initrange)
        init.constant_(self.v_embeddings.weight.data, 0) 

def reset_parameters(self):
        init.kaiming_normal_(self._weight, a=math.sqrt(5))
        fan_in, _ = init._calculate_fan_in_and_fan_out(self._weight)
        bound = 4 / math.sqrt(fan_in)
        init.uniform_(self._bias, -bound, bound)
        if self.over_param:
            with torch.no_grad(): self._bias.set_(self.manifold.expmap0(self._bias)) 

def reset_parameters(self):
        init.kaiming_uniform_(self.weight, a=math.sqrt(5))
        if self.bias is not None:
            fan_in, _ = init._calculate_fan_in_and_fan_out(self.weight)
            bound = 1 / math.sqrt(fan_in)
            init.uniform_(self.bias, -bound, bound) 

def mini_weight_init(m):
    if m.__class__.__name__ == 'Linear':
        m.weight.data.copy_(uniform_(m.weight.data, -3e-3, 3e-3))
        m.bias.data.fill_(0) 

def mini_weight_init(m):
    if m.__class__.__name__ == 'Linear':
        m.weight.data.copy_(uniform_(m.weight.data, -3e-3, 3e-3))
        m.bias.data.fill_(0) 

def mini_weight_init(m):
    if m.__class__.__name__ == 'Linear':
        m.weight.data.copy_(uniform_(m.weight.data, -3e-3, 3e-3))
        m.bias.data.fill_(0) 

def XavierFill(tensor):
    """Caffe2 XavierFill Implementation"""
    size = reduce(operator.mul, tensor.shape, 1)
    fan_in = size / tensor.shape[0]
    scale = math.sqrt(3 / fan_in)
    return init.uniform_(tensor, -scale, scale) 

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) 

def init_lstm_(lstm, init_weight=0.1):
    """
    Initializes weights of LSTM layer.
    Weights and biases are initialized with uniform(-init_weight, init_weight)
    distribution.

    :param lstm: instance of torch.nn.LSTM
    :param init_weight: range for the uniform initializer
    """
    # Initialize hidden-hidden weights
    init.uniform_(lstm.weight_hh_l0.data, -init_weight, init_weight)
    # Initialize input-hidden weights:
    init.uniform_(lstm.weight_ih_l0.data, -init_weight, init_weight)

    # Initialize bias. PyTorch LSTM has two biases, one for input-hidden GEMM
    # and the other for hidden-hidden GEMM. Here input-hidden bias is
    # initialized with uniform distribution and hidden-hidden bias is
    # initialized with zeros.
    init.uniform_(lstm.bias_ih_l0.data, -init_weight, init_weight)
    init.zeros_(lstm.bias_hh_l0.data)

    if lstm.bidirectional:
        init.uniform_(lstm.weight_hh_l0_reverse.data, -init_weight, init_weight)
        init.uniform_(lstm.weight_ih_l0_reverse.data, -init_weight, init_weight)

        init.uniform_(lstm.bias_ih_l0_reverse.data, -init_weight, init_weight)
        init.zeros_(lstm.bias_hh_l0_reverse.data) 

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) 

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) 

def XavierFill(tensor):
    """Caffe2 XavierFill Implementation"""
    size = reduce(operator.mul, tensor.shape, 1)
    fan_in = size / tensor.shape[0]
    scale = math.sqrt(3 / fan_in)
    return init.uniform_(tensor, -scale, scale) 

def reset_parameters(self):
        self.reset_running_stats()
        init.uniform_(self.weight)
        init.zeros_(self.bias) 

def reset_parameters(self):
        self.reset_running_stats()
        init.uniform_(self.weight)
        init.zeros_(self.bias) 

def __init__(self, input_dim, n_hidden, n_layer,
                 dropout, n_hop):
        super().__init__()
        self._init_h = nn.Parameter(torch.Tensor(n_layer, n_hidden))
        self._init_c = nn.Parameter(torch.Tensor(n_layer, n_hidden))
        self._init_i = nn.Parameter(torch.Tensor(input_dim))
        init.uniform_(self._init_h, -INI, INI)
        init.uniform_(self._init_c, -INI, INI)
        init.uniform_(self._init_i, -0.1, 0.1)
        self._lstm = nn.LSTM(
            input_dim, n_hidden, n_layer,
            bidirectional=False, dropout=dropout
        )
        self._lstm_cell = None

        # attention parameters
        self._attn_wm = nn.Parameter(torch.Tensor(input_dim, n_hidden))
        self._attn_wq = nn.Parameter(torch.Tensor(n_hidden, n_hidden))
        self._attn_v = nn.Parameter(torch.Tensor(n_hidden))
        init.xavier_normal_(self._attn_wm)
        init.xavier_normal_(self._attn_wq)
        init.uniform_(self._attn_v, -INI, INI)

        # hop parameters
        self._hop_wm = nn.Parameter(torch.Tensor(input_dim, n_hidden))
        self._hop_wq = nn.Parameter(torch.Tensor(n_hidden, n_hidden))
        self._hop_v = nn.Parameter(torch.Tensor(n_hidden))
        init.xavier_normal_(self._hop_wm)
        init.xavier_normal_(self._hop_wq)
        init.uniform_(self._hop_v, -INI, INI)
        self._n_hop = n_hop 

def __init__(self, input_dim, n_hidden, n_layer, dropout, bidirectional):
        super().__init__()
        self._init_h = nn.Parameter(
            torch.Tensor(n_layer*(2 if bidirectional else 1), n_hidden))
        self._init_c = nn.Parameter(
            torch.Tensor(n_layer*(2 if bidirectional else 1), n_hidden))
        init.uniform_(self._init_h, -INI, INI)
        init.uniform_(self._init_c, -INI, INI)
        self._lstm = nn.LSTM(input_dim, n_hidden, n_layer,
                             dropout=dropout, bidirectional=bidirectional) 

def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
        if args:
            ptr_net = args[0]
        else:
            ptr_net = kwargs['ptr_net']
        assert isinstance(ptr_net, LSTMPointerNet)
        self._stop = nn.Parameter(
            torch.Tensor(self._lstm_cell.input_size))
        init.uniform_(self._stop, -INI, INI) 

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) 

def XavierFill(tensor):
    """Caffe2 XavierFill Implementation"""
    size = reduce(operator.mul, tensor.shape, 1)
    fan_in = size / tensor.shape[0]
    scale = math.sqrt(3 / fan_in)
    return init.uniform_(tensor, -scale, scale) 

def reset_parameters(self):
        init.uniform_(self.init_param, - 1 / 2, 1 / 2) 

def reset_parameters(self):
        init.uniform_(self.initial_param, -math.sqrt(0.5), math.sqrt(0.5)) 

def reset_parameters(dim):
        init.uniform_(self.z0, -math.sqrt(1/dim), math.sqrt(1/dim))
        init.uniform_(self.log_alpha, -math.sqrt(1/dim), math.sqrt(1/dim))
        init.uniform_(self.beta, -math.sqrt(1/dim), math.sqrt(1/dim)) 

def reset_parameters(self, dim):
        init.uniform_(self.w, -math.sqrt(1/dim), math.sqrt(1/dim))
        init.uniform_(self.u, -math.sqrt(1/dim), math.sqrt(1/dim))
        init.uniform_(self.b, -math.sqrt(1/dim), math.sqrt(1/dim)) 

def __init__(self,args, batchNorm=True):
        super(FlowNetFusion,self).__init__()

        self.batchNorm = batchNorm
        self.conv0   = conv(self.batchNorm,  11,   64)
        self.conv1   = conv(self.batchNorm,  64,   64, stride=2)
        self.conv1_1 = conv(self.batchNorm,  64,   128)
        self.conv2   = conv(self.batchNorm,  128,  128, stride=2)
        self.conv2_1 = conv(self.batchNorm,  128,  128)

        self.deconv1 = deconv(128,32)
        self.deconv0 = deconv(162,16)

        self.inter_conv1 = i_conv(self.batchNorm,  162,   32)
        self.inter_conv0 = i_conv(self.batchNorm,  82,   16)

        self.predict_flow2 = predict_flow(128)
        self.predict_flow1 = predict_flow(32)
        self.predict_flow0 = predict_flow(16)

        self.upsampled_flow2_to_1 = nn.ConvTranspose2d(2, 2, 4, 2, 1)
        self.upsampled_flow1_to_0 = nn.ConvTranspose2d(2, 2, 4, 2, 1)

        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) 

def reset_parameters(self):
        self.reset_running_stats()
        init.uniform_(self.weight)
        init.zeros_(self.bias) 

def reset_parameters(self):
        self.reset_running_stats()
        init.uniform_(self.weight)
        init.zeros_(self.bias)