Ejemplo n.º 1
0
  def _init_weights(self, weight_shape, bias_shape):
    if self.initB is None:
      self.initB = 0.0

    if self.initW is None:
      self.initW = 1.0 / np.sqrt(np.prod(weight_shape))
     
    self.bias.shape = bias_shape
    self.weight.shape = weight_shape

    if self.weight.wt is None:
      self.weight.set_weight(to_gpu(col_randn(weight_shape, np.float32) * self.initW))

    if self.bias.wt is None:
      self.bias.set_weight(to_gpu((np.ones(bias_shape, dtype=np.float32) * self.initB)))
Ejemplo n.º 2
0
  def _init_weights(self, weight_shape, bias_shape):
    if self.initB is None:
      self.initB = 0.0

    if self.initW is None:
      self.initW = 1.0 / np.sqrt(np.prod(weight_shape))
     
    self.bias.shape = bias_shape
    self.weight.shape = weight_shape

    if self.weight.wt is None:
      self.weight.set_weight(to_gpu(col_randn(weight_shape, np.float32) * self.initW))

    if self.bias.wt is None:
      self.bias.set_weight(to_gpu((np.ones(bias_shape, dtype=np.float32) * self.initB)))
Ejemplo n.º 3
0
  def fprop(self, input, output, train=TRAIN):
    gpu_copy_to(dot(self.weight.wt, input), output)
    add_vec_to_rows(output, self.bias.wt)

    if train == TEST:
      if self.dropRate > 0.0:
        output *= (1.0 - self.dropRate)
    else:
      if self.dropRate > 0.0:
        self.dropMask = to_gpu(np.random.uniform(0, 1, output.size).astype(np.float32).reshape(output.shape))
        bigger_than_scaler(self.dropMask, self.dropRate)
        gpu_copy_to(output * self.dropMask, output)
    if PFout:
      print_matrix(output, self.name)
Ejemplo n.º 4
0
  def fprop(self, input, output, train=TRAIN):
    gpu_copy_to(dot(self.weight.wt, input), output)
    add_vec_to_rows(output, self.bias.wt)

    if train == TEST:
      if self.dropRate > 0.0:
        output *= (1.0 - self.dropRate)
    else:
      if self.dropRate > 0.0:
        self.dropMask = to_gpu(np.random.uniform(0, 1, output.size).astype(np.float32).reshape(output.shape))
        bigger_than_scaler(self.dropMask, self.dropRate)
        gpu_copy_to(output * self.dropMask, output)
    if PFout:
      print_matrix(output, self.name)
Ejemplo n.º 5
0
 def __init__(self, name, type, epsW, epsB, initW, initB, momW, momB, wc, weight, bias,
     weightIncr , biasIncr, disable_bprop=False):
   Layer.__init__(self, name, type, disable_bprop)
   self.initW = initW
   self.initB = initB
   
   self.weight = WEIGHTS.empty('weight.' + self.name, epsW, momW, wc)
   self.bias = WEIGHTS.empty('bias.' + self.name, epsB, momB, 0.0)
  
   if weight is not None:
     self.weight.set_weight(weight) 
   if weightIncr is not None:
     self.weight.set_incr(weightIncr)
   
   if bias is not None:
     self.bias.set_weight(bias) 
   if biasIncr is not None:
     self.bias.set_incr(to_gpu(biasIncr))
Ejemplo n.º 6
0
 def __init__(self, name, type, epsW, epsB, initW, initB, momW, momB, wc, weight, bias,
     weightIncr , biasIncr, disable_bprop=False):
   Layer.__init__(self, name, type, disable_bprop)
   self.initW = initW
   self.initB = initB
   
   self.weight = WEIGHTS.empty('weight.' + self.name, epsW, momW, wc)
   self.bias = WEIGHTS.empty('bias.' + self.name, epsB, momB, 0.0)
  
   if weight is not None:
     self.weight.set_weight(weight) 
   if weightIncr is not None:
     self.weight.set_incr(weightIncr)
   
   if bias is not None:
     self.bias.set_weight(bias) 
   if biasIncr is not None:
     self.bias.set_incr(to_gpu(biasIncr))