repo
stringlengths
2
99
file
stringlengths
13
225
code
stringlengths
0
18.3M
file_length
int64
0
18.3M
avg_line_length
float64
0
1.36M
max_line_length
int64
0
4.26M
extension_type
stringclasses
1 value
chainer
chainer-master/chainer/functions/pooling/roi_average_align_2d.py
# Modified work: # ----------------------------------------------------------------------------- # Copyright (c) 2018 Preferred Infrastructure, Inc. # Copyright (c) 2018 Preferred Networks, Inc. # ----------------------------------------------------------------------------- # Original work: # ----------------------------------------------------------------------------- # Copyright (c) 2015 by Contributors # \file roi_pooling.cu # \brief roi pooling operator # \author Ross Girshick, Kye-Hyeon Kim, Jian Guo # \changed to roi_align by Elaine Bao # \file roi_align.cu # \roi align operator described in Mask RCNN # ----------------------------------------------------------------------------- import numbers import numpy import six import chainer from chainer.backends import cuda from chainer import function from chainer import utils from chainer.utils import type_check def _pair(x): if isinstance(x, chainer.utils.collections_abc.Iterable): return x return x, x def _get_bounds(p, limit): if p < -1 or p > limit: # out of range, so it is empty return None, None, None low = int(numpy.floor(p)) if low == limit: low = low - 1 high = low + 1 if low <= -1: p = 0 elif high >= limit: p = limit - 1 return p, low, high def _get_bilinear_interp_params(y, x, y_low, x_low, y_high, x_high): ly = y - y_low lx = x - x_low hy = y_high - y hx = x_high - x w1 = hy * hx w2 = hy * lx w3 = ly * hx w4 = ly * lx return w1, w2, w3, w4 _GET_BILINEAR_INTERP_KERNEL = ''' __device__ bool get_bounds( T &p, const int limit, int &low, int &high) { if (p < -1. || p > limit) { // empty return false; } low = (int)floor(p); if (low == limit) { low = low - 1; } high = low + 1; if (low <= -1) { p = (T) 0.0; } else if (high >= limit) { p = (T) (limit - 1); } return true; } __device__ void get_bilinear_interp_params( T y, T x, int y_low, int x_low, int y_high, int x_high, T &w1, T &w2, T &w3, T &w4) { T ly = y - y_low; T lx = x - x_low; T hy = y_high - y; T hx = x_high - x; w1 = hy * hx; w2 = hy * lx; w3 = ly * hx; w4 = ly * lx; } ''' class ROIAverageAlign2D(function.Function): """ROI average align over a set of 2d planes.""" def __init__(self, outsize, spatial_scale, sampling_ratio=None): outh, outw = _pair(outsize) if not (isinstance(outh, numbers.Integral) and outh > 0): raise TypeError( 'outsize[0] must be positive integer: {}, {}' .format(type(outh), outh)) if not (isinstance(outw, numbers.Integral) and outw > 0): raise TypeError( 'outsize[1] must be positive integer: {}, {}' .format(type(outw), outw)) if isinstance(spatial_scale, numbers.Integral): spatial_scale = float(spatial_scale) if not (isinstance(spatial_scale, numbers.Real) and spatial_scale > 0): raise TypeError( 'spatial_scale must be a positive float number: {}, {}' .format(type(spatial_scale), spatial_scale)) sampling_ratio = _pair(sampling_ratio) if not all((isinstance(s, numbers.Integral) and s >= 1) or s is None for s in sampling_ratio): raise TypeError( 'sampling_ratio must be integer >= 1 or a pair of it: {}' .format(sampling_ratio)) self.outh, self.outw = outh, outw self.spatial_scale = spatial_scale self.sampling_ratio = sampling_ratio def check_type_forward(self, in_types): type_check.expect(in_types.size() == 3) x_type, roi_type, roi_index_type = in_types type_check.expect( x_type.dtype == numpy.float32, x_type.ndim == 4, roi_type.dtype == numpy.float32, roi_type.ndim == 2, roi_type.shape[1] == 4, roi_index_type.dtype == numpy.int32, roi_index_type.ndim == 1, roi_type.shape[0] == roi_index_type.shape[0], ) def forward_cpu(self, inputs): self.retain_inputs((1, 2)) self._bottom_data_shape = inputs[0].shape bottom_data, bottom_rois, bottom_roi_indices = inputs channels, height, width = bottom_data.shape[1:] n_rois = bottom_rois.shape[0] top_data = numpy.empty((n_rois, channels, self.outh, self.outw), dtype=bottom_data.dtype) pooled_width, pooled_height = self.outw, self.outh spatial_scale = self.spatial_scale for i in six.moves.range(top_data.size): pw = i % pooled_width ph = int(i / pooled_width) % pooled_height c = int(i / pooled_width / pooled_height) % channels n = int(i / pooled_width / pooled_height / channels) roi_batch_ind = int(bottom_roi_indices[n]) roi_start_h = bottom_rois[n, 0] * spatial_scale roi_start_w = bottom_rois[n, 1] * spatial_scale roi_end_h = bottom_rois[n, 2] * spatial_scale roi_end_w = bottom_rois[n, 3] * spatial_scale roi_height = max(roi_end_h - roi_start_h, 1.) roi_width = max(roi_end_w - roi_start_w, 1.) bin_size_h = roi_height / pooled_height bin_size_w = roi_width / pooled_width if self.sampling_ratio[0] is None: roi_bin_grid_h = int(numpy.ceil(roi_height / pooled_height)) else: roi_bin_grid_h = self.sampling_ratio[0] if self.sampling_ratio[1] is None: roi_bin_grid_w = int(numpy.ceil(roi_width / pooled_width)) else: roi_bin_grid_w = self.sampling_ratio[1] count = roi_bin_grid_h * roi_bin_grid_w output_val = 0. for iy in six.moves.range(roi_bin_grid_h): y = roi_start_h + ph * bin_size_h + \ (iy + .5) * bin_size_h / roi_bin_grid_h y, y_low, y_high = _get_bounds(y, height) if y is None or y_low is None or y_high is None: continue for ix in six.moves.range(roi_bin_grid_w): x = roi_start_w + pw * bin_size_w + \ (ix + .5) * bin_size_w / roi_bin_grid_w x, x_low, x_high = _get_bounds(x, width) if x is None or x_low is None or x_high is None: continue # bilinear interpolation {{ w1, w2, w3, w4 = _get_bilinear_interp_params( y, x, y_low, x_low, y_high, x_high) if w1 > 0 and y_low >= 0 and x_low >= 0: v1 = bottom_data[roi_batch_ind, c, y_low, x_low] output_val += w1 * v1 if w2 > 0 and y_low >= 0 and x_high <= width - 1: v2 = bottom_data[roi_batch_ind, c, y_low, x_high] output_val += w2 * v2 if w3 > 0 and y_high <= height - 1 and x_low >= 0: v3 = bottom_data[roi_batch_ind, c, y_high, x_low] output_val += w3 * v3 if w4 > 0 and y_high <= height - 1 and x_high <= width - 1: v4 = bottom_data[roi_batch_ind, c, y_high, x_high] output_val += w4 * v4 # }} output_val /= count top_data[n, c, ph, pw] = output_val return top_data, def forward_gpu(self, inputs): self.retain_inputs((1, 2)) self._bottom_data_shape = inputs[0].shape bottom_data, bottom_rois, bottom_roi_indices = inputs channels, height, width = bottom_data.shape[1:] n_rois = bottom_rois.shape[0] top_data = cuda.cupy.empty((n_rois, channels, self.outh, self.outw), dtype=bottom_data.dtype) if self.sampling_ratio[0] is None: sampling_ratio_h = 0 else: sampling_ratio_h = self.sampling_ratio[0] if self.sampling_ratio[1] is None: sampling_ratio_w = 0 else: sampling_ratio_w = self.sampling_ratio[1] cuda.elementwise( ''' raw T bottom_data, T spatial_scale, int32 channels, int32 height, int32 width, int32 pooled_height, int32 pooled_width, int32 sampling_ratio_h, int32 sampling_ratio_w, raw T bottom_rois, raw int32 bottom_roi_indices ''', 'T top_data', ''' int pw = i % pooled_width; int ph = (i / pooled_width) % pooled_height; int c = (i / pooled_width / pooled_height) % channels; int n = i / pooled_width / pooled_height / channels; int roi_batch_ind = bottom_roi_indices[n]; T roi_start_h = bottom_rois[n * 4 + 0] * spatial_scale; T roi_start_w = bottom_rois[n * 4 + 1] * spatial_scale; T roi_end_h = bottom_rois[n * 4 + 2] * spatial_scale; T roi_end_w = bottom_rois[n * 4 + 3] * spatial_scale; // Force malformed ROIs to be 1x1 T roi_height = max(roi_end_h - roi_start_h, (T)1.); T roi_width = max(roi_end_w - roi_start_w, (T)1.); T bin_size_h = static_cast<T>(roi_height) / static_cast<T>(pooled_height); T bin_size_w = static_cast<T>(roi_width) / static_cast<T>(pooled_width); int bottom_data_offset = (roi_batch_ind * channels + c) * height * width; // We use roi_bin_grid to sample the grid and mimic integral int roi_bin_grid_h = (sampling_ratio_h > 0) ? sampling_ratio_h : ceil(roi_height / pooled_height); // e.g. = 2 int roi_bin_grid_w = (sampling_ratio_w > 0) ? sampling_ratio_w : ceil(roi_width / pooled_width); // We do average (integral) pooling inside a bin T count = roi_bin_grid_h * roi_bin_grid_w; // e.g. = 4 T output_val = 0.; for (int iy = 0; iy < roi_bin_grid_h; iy++) // e.g. iy = 0, 1 { T y = roi_start_h + ph * bin_size_h + static_cast<T>(iy + .5f) * bin_size_h / static_cast<T>(roi_bin_grid_h); // e.g. 0.5, 1.5 int y_low, y_high; bool y_ret = get_bounds(y, height, y_low, y_high); if (!y_ret) continue; for (int ix = 0; ix < roi_bin_grid_w; ix++) { T x = roi_start_w + pw * bin_size_w + static_cast<T>(ix + .5f) * bin_size_w / static_cast<T>(roi_bin_grid_w); int x_low, x_high; bool x_ret = get_bounds(x, width, x_low, x_high); if (!x_ret) continue; // bilinear_interpolation_gradient {{ T w1, w2, w3, w4; get_bilinear_interp_params( y, x, y_low, x_low, y_high, x_high, w1, w2, w3, w4); if (w1 > 0 && y_low >= 0 && x_low >= 0) { T v1 = bottom_data[ bottom_data_offset + y_low * width + x_low]; output_val += w1 * v1; } if (w2 > 0 && y_low >= 0 && x_high <= width - 1) { T v2 = bottom_data[ bottom_data_offset + y_low * width + x_high]; output_val += w2 * v2; } if (w3 > 0 && y_high <= height - 1 && x_low >= 0) { T v3 = bottom_data[ bottom_data_offset + y_high * width + x_low]; output_val += w3 * v3; } if (w4 > 0 && y_high <= height - 1 && x_high <= width - 1) { T v4 = bottom_data[ bottom_data_offset + y_high * width + x_high]; output_val += w4 * v4; } // }} } } output_val /= count; top_data = output_val; ''', 'roi_average_align_2d_fwd', preamble=_GET_BILINEAR_INTERP_KERNEL, )(bottom_data, self.spatial_scale, channels, height, width, self.outh, self.outw, sampling_ratio_h, sampling_ratio_w, bottom_rois, bottom_roi_indices, top_data) return top_data, def backward_cpu(self, inputs, gy): bottom_rois, bottom_roi_indices = inputs[1:] channels, height, width = self._bottom_data_shape[1:] bottom_diff = numpy.zeros(self._bottom_data_shape, gy[0].dtype) spatial_scale = self.spatial_scale pooled_height = self.outh pooled_width = self.outw top_diff = gy[0] for i in six.moves.range(top_diff.size): pw = i % pooled_width ph = int(i / pooled_width) % pooled_height c = int(i / pooled_width / pooled_height) % channels n = int(i / pooled_width / pooled_height / channels) roi_batch_ind = int(bottom_roi_indices[n]) roi_start_h = bottom_rois[n, 0] * spatial_scale roi_start_w = bottom_rois[n, 1] * spatial_scale roi_end_h = bottom_rois[n, 2] * spatial_scale roi_end_w = bottom_rois[n, 3] * spatial_scale roi_height = max(roi_end_h - roi_start_h, 1.) roi_width = max(roi_end_w - roi_start_w, 1.) bin_size_h = roi_height / pooled_height bin_size_w = roi_width / pooled_width top_diff_this_bin = top_diff[n, c, ph, pw] if self.sampling_ratio[0] is None: roi_bin_grid_h = int(numpy.ceil(roi_height / pooled_height)) else: roi_bin_grid_h = self.sampling_ratio[0] if self.sampling_ratio[1] is None: roi_bin_grid_w = int(numpy.ceil(roi_width / pooled_width)) else: roi_bin_grid_w = self.sampling_ratio[1] count = roi_bin_grid_h * roi_bin_grid_w for iy in six.moves.range(roi_bin_grid_h): y = roi_start_h + ph * bin_size_h + \ (iy + .5) * bin_size_h / roi_bin_grid_h y, y_low, y_high = _get_bounds(y, height) if y is None or y_low is None or y_high is None: continue for ix in six.moves.range(roi_bin_grid_w): x = roi_start_w + pw * bin_size_w + \ (ix + .5) * bin_size_w / roi_bin_grid_w x, x_low, x_high = _get_bounds(x, width) if x is None or x_low is None or x_high is None: continue # bilinear_interpolation_gradient {{ w1, w2, w3, w4 = _get_bilinear_interp_params( y, x, y_low, x_low, y_high, x_high) if w1 > 0 and y_low >= 0 and x_low >= 0: g1 = top_diff_this_bin * w1 / count bottom_diff[roi_batch_ind, c, y_low, x_low] += g1 if w2 > 0 and y_low >= 0 and x_high <= width - 1: g2 = top_diff_this_bin * w2 / count bottom_diff[roi_batch_ind, c, y_low, x_high] += g2 if w3 > 0 and y_high <= height - 1 and x_low >= 0: g3 = top_diff_this_bin * w3 / count bottom_diff[roi_batch_ind, c, y_high, x_low] += g3 if w4 > 0 and y_high <= height - 1 and x_high <= width - 1: g4 = top_diff_this_bin * w4 / count bottom_diff[roi_batch_ind, c, y_high, x_high] += g4 # }} return bottom_diff, None, None def backward_gpu(self, inputs, gy): utils.nondeterministic('atomicAdd') bottom_rois, bottom_roi_indices = inputs[1:] channels, height, width = self._bottom_data_shape[1:] bottom_diff = cuda.cupy.zeros(self._bottom_data_shape, gy[0].dtype) if self.sampling_ratio[0] is None: sampling_ratio_h = 0 else: sampling_ratio_h = self.sampling_ratio[0] if self.sampling_ratio[1] is None: sampling_ratio_w = 0 else: sampling_ratio_w = self.sampling_ratio[1] cuda.elementwise( ''' raw T top_diff, int32 num_rois, T spatial_scale, int32 channels, int32 height, int32 width, int32 pooled_height, int32 pooled_width, int32 sampling_ratio_h, int32 sampling_ratio_w, raw T bottom_rois, raw int32 bottom_roi_indices ''', 'raw T bottom_diff', ''' // (n, c, h, w) coords in bottom data int pw = i % pooled_width; int ph = (i / pooled_width) % pooled_height; int c = (i / pooled_width / pooled_height) % channels; int n = i / pooled_width / pooled_height / channels; // Do not using rounding; this implementation detail is critical int roi_batch_ind = bottom_roi_indices[n]; T roi_start_h = bottom_rois[n * 4 + 0] * spatial_scale; T roi_start_w = bottom_rois[n * 4 + 1] * spatial_scale; T roi_end_h = bottom_rois[n * 4 + 2] * spatial_scale; T roi_end_w = bottom_rois[n * 4 + 3] * spatial_scale; // Force malformed ROIs to be 1x1 T roi_width = max(roi_end_w - roi_start_w, (T)1.); T roi_height = max(roi_end_h - roi_start_h, (T)1.); T bin_size_h = static_cast<T>(roi_height) / static_cast<T>(pooled_height); T bin_size_w = static_cast<T>(roi_width) / static_cast<T>(pooled_width); int bottom_diff_offset = (roi_batch_ind * channels + c) * height * width; int top_offset = (n * channels + c) * pooled_height * pooled_width; T top_diff_this_bin = top_diff[top_offset + ph * pooled_width + pw]; // We use roi_bin_grid to sample the grid and mimic integral int roi_bin_grid_h = (sampling_ratio_h > 0) ? sampling_ratio_h : ceil(roi_height / pooled_height); // e.g. = 2 int roi_bin_grid_w = (sampling_ratio_w > 0) ? sampling_ratio_w : ceil(roi_width / pooled_width); // We do average (integral) pooling inside a bin T count = roi_bin_grid_h * roi_bin_grid_w; // e.g. = 4 for (int iy = 0; iy < roi_bin_grid_h; iy++) { T y = roi_start_h + ph * bin_size_h + static_cast<T>(iy + .5f) * bin_size_h / static_cast<T>(roi_bin_grid_h); // e.g. 0.5, 1.5 int y_low, y_high; bool y_ret = get_bounds(y, height, y_low, y_high); if (!y_ret) continue; for (int ix = 0; ix < roi_bin_grid_w; ix++) { T x = roi_start_w + pw * bin_size_w + static_cast<T>(ix + .5f) * bin_size_w / static_cast<T>(roi_bin_grid_w); int x_low, x_high; bool x_ret = get_bounds(x, width, x_low, x_high); if (!x_ret) continue; // bilinear_interpolation_gradient {{ T w1, w2, w3, w4; get_bilinear_interp_params( y, x, y_low, x_low, y_high, x_high, w1, w2, w3, w4); if (w1 > 0 && y_low >= 0 && x_low >= 0) { T g1 = top_diff_this_bin * w1 / count; atomicAdd(&bottom_diff[ bottom_diff_offset + y_low * width + x_low], g1); } if (w2 > 0 && y_low >= 0 && x_high <= width - 1) { T g2 = top_diff_this_bin * w2 / count; atomicAdd(&bottom_diff[ bottom_diff_offset + y_low * width + x_high], g2); } if (w3 > 0 && y_high <= height - 1 && x_low >= 0) { T g3 = top_diff_this_bin * w3 / count; atomicAdd(&bottom_diff[ bottom_diff_offset + y_high * width + x_low], g3); } if (w4 > 0 && y_high <= height - 1 && x_high <= width - 1) { T g4 = top_diff_this_bin * w4 / count; atomicAdd(&bottom_diff[ bottom_diff_offset + y_high * width + x_high], g4); } // }} } } ''', 'roi_average_align_2d_bwd', preamble=_GET_BILINEAR_INTERP_KERNEL, )(gy[0], bottom_rois.shape[0], self.spatial_scale, channels, height, width, self.outh, self.outw, sampling_ratio_h, sampling_ratio_w, bottom_rois, bottom_roi_indices, bottom_diff, size=gy[0].size) return bottom_diff, None, None def roi_average_align_2d( x, rois, roi_indices, outsize, spatial_scale, sampling_ratio=None ): """Spatial Region of Interest (ROI) average align function. This function acts similarly to :func:`~chainer.functions.roi_average_pooling_2d`, but it computes average of input spatial patch with bilinear interpolation for each channel with the region of interest. Args: x (~chainer.Variable): Input variable. The shape is expected to be 4 dimensional: ``(n: batch, c: channel, h, height, w: width)``. rois (~chainer.Variable): Input roi variable. The shape is expected to be ``(n: data size, 4)``, and each datum is set as below: ``(y_min, x_min, y_max, x_max)``. roi_indices (~chainer.Variable): Input roi variable. The shape is expected to be ``(n: data size, )``. outsize ((int, int) or int): Expected output size after pooled (height, width). ``outsize=o`` and ``outsize=(o, o)`` are equivalent. spatial_scale (float): Scale of the roi is resized. sampling_ratio ((int, int) or int): Sampling step for the alignment. It must be an integer over :math:`1` or :obj:`None`, and the value is automatically decided when :obj:`None` is passed. Use of different ratio in height and width axis is also supported by passing tuple of int as ``(sampling_ratio_h, sampling_ratio_w)``. ``sampling_ratio=s`` and ``sampling_ratio=(s, s)`` are equivalent. Returns: ~chainer.Variable: Output variable. See the original paper proposing ROIAlign: `Mask R-CNN <https://arxiv.org/abs/1703.06870>`_. """ return ROIAverageAlign2D(outsize, spatial_scale, sampling_ratio)( x, rois, roi_indices)
23,553
39.332192
79
py
chainer
chainer-master/chainer/functions/pooling/average_pooling_2d.py
import numpy import chainer from chainer import backend from chainer.backends import cuda from chainer.backends import intel64 from chainer import function_node from chainer.functions.pooling import average_pooling_nd from chainer.functions.pooling import pooling_2d from chainer.utils import conv import chainerx class AveragePooling2D(pooling_2d.Pooling2D): """Average pooling over a set of 2d planes.""" # TODO(beam2d): Support cover_all mode. def forward_cpu(self, x): if (intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(x)): return self._forward_ideep(x) self._in_shape = x[0].shape self._in_dtype = x[0].dtype col = conv.im2col_cpu(x[0], self.kh, self.kw, self.sy, self.sx, self.ph, self.pw) y = col.mean(axis=(2, 3)) return y, def _forward_ideep(self, x): self._in_shape = x[0].shape self._in_dtype = x[0].dtype self.retain_inputs((0,)) n, c, h, w = x[0].shape y_h = conv.get_conv_outsize( h, self.kh, self.sy, self.ph, self.cover_all) assert y_h > 0, 'Height in the output should be positive.' y_w = conv.get_conv_outsize( w, self.kw, self.sx, self.pw, self.cover_all) assert y_w > 0, 'Width in the output should be positive.' pd = self.sy * (y_h - 1) + self.kh - h - self.ph pr = self.sx * (y_w - 1) + self.kw - w - self.pw pp = intel64.ideep.pooling2DParam( (n, c, y_h, y_w), self.kh, self.kw, self.sy, self.sx, self.ph, self.pw, pd, pr, intel64.ideep.pooling2DParam.pooling_avg_include_padding) y, = intel64.ideep.pooling2D.Forward(intel64.ideep.array(x[0]), pp) return y, def forward_gpu(self, x): if chainer.should_use_cudnn('>=auto'): self.retain_inputs((0,)) return super(AveragePooling2D, self).forward_gpu(x) self._in_shape = x[0].shape self._in_dtype = x[0].dtype n, c, h, w = x[0].shape y_h = conv.get_conv_outsize(h, self.kh, self.sy, self.ph) y_w = conv.get_conv_outsize(w, self.kw, self.sx, self.pw) y = cuda.cupy.empty((n, c, y_h, y_w), dtype=x[0].dtype) coeff = 1. / (self.kh * self.kw) kern = cuda.elementwise( 'raw T in, int32 h, int32 w,' 'int32 out_h, int32 out_w, int32 kh, int32 kw,' 'int32 sy, int32 sx, int32 ph, int32 pw, T coeff', 'T out', ''' int c0 = i / (out_h * out_w); int out_y = i / out_w % out_h; int out_x = i % out_w; int in_y_0 = max(0, out_y * sy - ph); int in_y_1 = min(h, out_y * sy + kh - ph); int in_x_0 = max(0, out_x * sx - pw); int in_x_1 = min(w, out_x * sx + kw - pw); T val = 0; for (int y = in_y_0; y < in_y_1; ++y) { int offset_y = w * (y + h * c0); for (int x = in_x_0; x < in_x_1; ++x) { val = val + in[x + offset_y]; } } out = val * coeff; ''', 'avg_pool_fwd') kern(x[0].reduced_view(), h, w, y_h, y_w, self.kh, self.kw, self.sy, self.sx, self.ph, self.pw, coeff, y) return y, def backward(self, indexes, gy): return AveragePooling2DGrad(self).apply(gy) def _get_pool_mode(self): return cuda.cuda.cudnn.CUDNN_POOLING_AVERAGE_COUNT_INCLUDE_PADDING class AveragePooling2DGrad(function_node.FunctionNode): def __init__(self, apool2d): self.kh = apool2d.kh self.kw = apool2d.kw self.sy = apool2d.sy self.sx = apool2d.sx self.ph = apool2d.ph self.pw = apool2d.pw self._used_cudnn = apool2d._used_cudnn if not self._used_cudnn: self._in_shape = apool2d._in_shape self._in_dtype = apool2d._in_dtype self.apool2d = apool2d def forward_cpu(self, gy): if (intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(gy)): return self._forward_ideep(gy) h, w = self._in_shape[2:] gcol = numpy.tile(gy[0][:, :, None, None], (1, 1, self.kh, self.kw, 1, 1)) gx = conv.col2im_cpu(gcol, self.sy, self.sx, self.ph, self.pw, h, w) gx /= self.kh * self.kw return gx, def _forward_ideep(self, gy): n, c, h, w = self._in_shape y_h, y_w = gy[0].shape[2:] x, = self.apool2d.get_retained_inputs() pd = self.sy * (y_h - 1) + self.kh - h - self.ph pr = self.sx * (y_w - 1) + self.kw - w - self.pw pp = intel64.ideep.pooling2DParam( self._in_shape, self.kh, self.kw, self.sy, self.sx, self.ph, self.pw, pd, pr, intel64.ideep.pooling2DParam.pooling_avg_include_padding) gx = intel64.ideep.pooling2D.Backward( intel64.ideep.array(x.data), intel64.ideep.array(gy[0]), None, pp) return gx, def forward_gpu(self, gy): if self._used_cudnn: x, = self.apool2d.get_retained_inputs() return self.apool2d.backward_gpu((x.data,), gy) n, c, h, w = self._in_shape y_h, y_w = gy[0].shape[2:] gx = cuda.cupy.empty(self._in_shape, self._in_dtype) coeff = 1. / (self.kh * self.kw) cuda.elementwise( 'raw T gy, int32 h, int32 w,' 'int32 out_h, int32 out_w, int32 kh, int32 kw,' 'int32 sy, int32 sx, int32 ph, int32 pw, T coeff', 'T gx', ''' int c0 = i / (h * w); int y = i / w % h + ph; int x = i % w + pw; int out_y_0 = max(0, (y - kh + sy) / sy); int out_y_1 = min(out_h, (y + sy) / sy); int out_x_0 = max(0, (x - kw + sx) / sx); int out_x_1 = min(out_w, (x + sx) / sx); int hc0 = out_h * c0; T val = 0; for (int out_y = out_y_0; out_y < out_y_1; ++out_y) { for (int out_x = out_x_0; out_x < out_x_1; ++out_x) { val = val + gy[out_x + out_w * (out_y + hc0)]; } } gx = val * coeff; ''', 'avg_pool_bwd')(gy[0].reduced_view(), h, w, y_h, y_w, self.kh, self.kw, self.sy, self.sx, self.ph, self.pw, coeff, gx) return gx, def backward(self, indexes, grad_outputs): return AveragePooling2D( (self.kh, self.kw), (self.sy, self.sx), (self.ph, self.pw), False).apply(grad_outputs) def average_pooling_2d(x, ksize, stride=None, pad=0): """Spatial average pooling function. This function acts similarly to :func:`~chainer.functions.convolution_2d`, but it computes the average of input spatial patch for each channel without any parameter instead of computing the inner products. Args: x (~chainer.Variable): Input variable. ksize (int or pair of ints): Size of pooling window. ``ksize=k`` and ``ksize=(k, k)`` are equivalent. stride (int or pair of ints or None): Stride of pooling applications. ``stride=s`` and ``stride=(s, s)`` are equivalent. If ``None`` is specified, then it uses same stride as the pooling window size. pad (int or pair of ints): Spatial padding width for the input array. ``pad=p`` and ``pad=(p, p)`` are equivalent. Returns: ~chainer.Variable: Output variable. .. note:: This function currently does not support ``cover_all`` mode as :func:`max_pooling_2d`. Average pooling runs in non-cover-all mode. .. note:: The values in the padded region is treated as 0, leading the averages biased towards zero. To obtain unbiased averages, use :func:`average_pooling_nd` with ``pad_value=None``. """ if backend.get_array_module(x) is chainerx: return average_pooling_nd.average_pooling_nd(x, ksize, stride, pad) return AveragePooling2D(ksize, stride, pad, False).apply((x,))[0]
8,383
35.77193
79
py
chainer
chainer-master/chainer/functions/pooling/upsampling_2d.py
import numpy from chainer.backends import cuda from chainer import function_node from chainer.functions.pooling import pooling_2d from chainer.utils import conv from chainer.utils import type_check class Upsampling2D(pooling_2d.Pooling2D): """Upsampling over a set of 2d planes w/ indices used for max pooling.""" def __init__(self, indexes, ksize, stride=None, pad=0, outsize=None, cover_all=True): super(Upsampling2D, self).__init__(ksize, stride, pad, cover_all) self.indexes = indexes self.outh, self.outw = (None, None) if outsize is None else outsize def check_type_forward(self, in_types): n_in = in_types.size() type_check.expect(n_in == 1) x_type = in_types[0] type_check.expect( x_type.dtype.kind == 'f', x_type.ndim == 4, x_type.shape == self.indexes.shape, ) if self.outh is not None: expected_h = conv.get_conv_outsize( self.outh, self.kh, self.sy, self.ph, cover_all=self.cover_all) type_check.expect(x_type.shape[2] == expected_h) if self.outw is not None: expected_w = conv.get_conv_outsize( self.outw, self.kw, self.sx, self.pw, cover_all=self.cover_all) type_check.expect(x_type.shape[3] == expected_w) def forward_cpu(self, x): self._in_dtype = x[0].dtype n, c, h, w = x[0].shape if self.outh is None: self.outh = conv.get_deconv_outsize( h, self.kh, self.sy, self.ph, cover_all=self.cover_all) if self.outw is None: self.outw = conv.get_deconv_outsize( w, self.kw, self.sx, self.pw, cover_all=self.cover_all) up_y = numpy.zeros((n, c, self.outh, self.outw), dtype=self._in_dtype) up_y = conv.im2col_cpu( up_y, self.kh, self.kw, self.sy, self.sx, self.ph, self.pw, cover_all=self.cover_all).transpose(0, 1, 4, 5, 2, 3) colh, colw = up_y.shape[2:4] up_y = up_y.reshape(-1, self.kh * self.kw) indexes = self.indexes.ravel() up_y[numpy.arange(len(indexes)), indexes] = x[0].ravel() up_y = up_y.reshape(n, c, colh, colw, self.kh, self.kw) up_y = conv.col2im_cpu( up_y.transpose(0, 1, 4, 5, 2, 3), self.sy, self.sx, self.ph, self.pw, self.outh, self.outw) return up_y, def forward_gpu(self, x): self._in_dtype = x[0].dtype xp = cuda.cupy n, c, h, w = x[0].shape if self.outh is None: self.outh = conv.get_deconv_outsize( h, self.kh, self.sy, self.ph, cover_all=self.cover_all) if self.outw is None: self.outw = conv.get_deconv_outsize( w, self.kw, self.sx, self.pw, cover_all=self.cover_all) up_y = xp.zeros((n, c, self.outh, self.outw), dtype=self._in_dtype) up_y = conv.im2col_gpu( up_y, self.kh, self.kw, self.sy, self.sx, self.ph, self.pw, cover_all=self.cover_all) up_y = up_y.transpose(0, 1, 4, 5, 2, 3) n, c, oy, ox, ky, kx = up_y.shape indexes = xp.asarray(self.indexes, dtype=numpy.int32) cuda.elementwise( 'int32 index, T x, int32 n, int32 c, int32 oy, int32 ox,' 'int32 ky, int32 kx', 'raw T up_y', ''' int yn = i / c / oy / ox; int yc = (i / oy / ox) % c; int yoy = (i / ox) % oy; int yox = i % ox; up_y[yn * c * oy * ox * ky * kx + yc * oy * ox * ky * kx + yoy * ox * ky * kx + yox * ky * kx + index] = x; ''', 'upsampling_2d_fwd')(indexes, x[0], n, c, oy, ox, ky, kx, up_y) up_y = up_y.transpose(0, 1, 4, 5, 2, 3) up_y = conv.col2im_gpu(up_y, self.sy, self.sx, self.ph, self.pw, self.outh, self.outw) return up_y, def backward(self, indexes, grad_outputs): return Upsampling2DGrad(self).apply(grad_outputs) class Upsampling2DGrad(function_node.FunctionNode): def __init__(self, upsampling2d): self.kh = upsampling2d.kh self.kw = upsampling2d.kw self.sy = upsampling2d.sy self.sx = upsampling2d.sx self.ph = upsampling2d.ph self.pw = upsampling2d.pw self.outh = upsampling2d.outh self.outw = upsampling2d.outw self.cover_all = upsampling2d.cover_all self.indexes = upsampling2d.indexes self._in_dtype = upsampling2d._in_dtype def forward_cpu(self, gy): gcol = conv.im2col_cpu( gy[0], self.kh, self.kw, self.sy, self.sx, self.ph, self.pw, cover_all=self.cover_all) n, c, kh, kw, out_h, out_w = gcol.shape gcol = gcol.transpose(0, 1, 4, 5, 2, 3).reshape(-1, kh * kw) indexes = self.indexes.ravel() gx = gcol[numpy.arange(len(indexes)), indexes] return gx.reshape(n, c, out_h, out_w), def forward_gpu(self, gy): xp = cuda.cupy gcol = conv.im2col_gpu( gy[0], self.kh, self.kw, self.sy, self.sx, self.ph, self.pw, cover_all=self.cover_all) gcol = gcol.transpose(0, 1, 4, 5, 2, 3) n, c, oy, ox, ky, kx = gcol.shape gcol = gcol.reshape((n, c, oy, ox, ky * kx)) indexes = xp.asarray(self.indexes, dtype=numpy.int32) gx = xp.empty((n, c, oy, ox), dtype=self._in_dtype) cuda.elementwise( 'int32 indexes, raw T gcol, int32 n, int32 c, int32 oy,' 'int32 ox, int32 ky, int32 kx', 'raw T gx', ''' int ind_n = i / c / oy / ox; int ind_c = (i / oy / ox) % c; int ind_oy = (i / ox) % oy; int ind_ox = i % ox; int gcol_ky = indexes / kx; int gcol_kx = indexes % kx; float top_gx = gcol[ind_n * c * oy * ox * ky * kx + ind_c * oy * ox * ky * kx + ind_oy * ox * ky * kx + ind_ox * ky * kx + gcol_ky * kx + gcol_kx]; gx[ind_n * c * oy * ox + ind_c * oy * ox + ind_oy * ox + ind_ox] = top_gx; ''', 'upsampling_2d_bwd')(indexes, gcol, n, c, oy, ox, ky, kx, gx) return gx, def backward(self, indexes, ggx): return Upsampling2D( self.indexes, (self.kh, self.kw), (self.sy, self.sx), (self.ph, self.pw), (self.outh, self.outw), self.cover_all).apply(ggx) def upsampling_2d( x, indexes, ksize, stride=None, pad=0, outsize=None, cover_all=True): """Upsampling using pooling indices. This function produces an upsampled image using pooling indices. .. admonition:: Example >>> x = np.arange(1, 37).reshape(1, 1, 6, 6).astype(np.float32) >>> x = chainer.Variable(x) >>> x.array array([[[[ 1., 2., 3., 4., 5., 6.], [ 7., 8., 9., 10., 11., 12.], [13., 14., 15., 16., 17., 18.], [19., 20., 21., 22., 23., 24.], [25., 26., 27., 28., 29., 30.], [31., 32., 33., 34., 35., 36.]]]], dtype=float32) This is the original ``x`` before max pooling. >>> pooled_x, indexes = F.max_pooling_2d( ... x, ksize=2, stride=2, return_indices=True) >>> pooled_x.array array([[[[ 8., 10., 12.], [20., 22., 24.], [32., 34., 36.]]]], dtype=float32) >>> indexes array([[[[3, 3, 3], [3, 3, 3], [3, 3, 3]]]]) These are the outputs from the max pooling operation including the resulting indices that will be used to upsample ``pooled_x``. Note that the indices all point to the largest, in the case the last, elements in each window. >>> upsampled_x = F.upsampling_2d( ... pooled_x, indexes, ksize=2, stride=2, outsize=x.shape[2:]) >>> upsampled_x.shape (1, 1, 6, 6) >>> upsampled_x.array array([[[[ 0., 0., 0., 0., 0., 0.], [ 0., 8., 0., 10., 0., 12.], [ 0., 0., 0., 0., 0., 0.], [ 0., 20., 0., 22., 0., 24.], [ 0., 0., 0., 0., 0., 0.], [ 0., 32., 0., 34., 0., 36.]]]], dtype=float32) Args: x (~chainer.Variable): Input variable. indexes (:ref:`ndarray`): Index array returned from preceding call to :meth:`~chainer.functions.max_pooling_2d`. ksize (int or pair of ints): Size of pooling window. ``ksize=k`` and ``ksize=(k, k)`` are equivalent. stride (int or pair of ints or None): Stride of pooling applications. ``stride=s`` and ``stride=(s, s)`` are equivalent. If ``None`` is specified, then it uses same stride as the pooling window size. pad (int or pair of ints): Spatial padding width for the input array. ``pad=p`` and ``pad=(p, p)`` are equivalent. outsize ((int, int)): Expected output size (height, width). cover_all (bool): Should be set to ``True`` if all spatial locations were pooled into some output pixels during the preceding pooling operation. ``False`` otherwise. See :meth:`~chainer.functions.max_pooling_2d`. Returns: ~chainer.Variable: Output variable. """ return Upsampling2D( indexes, ksize, stride, pad, outsize, cover_all).apply((x,))[0]
9,757
38.666667
79
py
chainer
chainer-master/chainer/functions/pooling/unpooling_nd.py
import numpy import six from chainer import backend from chainer import function_node from chainer.functions.pooling import pooling_nd from chainer.utils import conv from chainer.utils import conv_nd from chainer.utils import type_check class UnpoolingND(pooling_nd._PoolingND): """Unpooling over a set of N-dimensional planes. .. warning:: This feature is experimental. The interface can change in the future. """ def __init__(self, ndim, ksize, stride=None, pad=0, outsize=None, cover_all=True): super(UnpoolingND, self).__init__(ndim, ksize, stride, pad, cover_all) self.outs = None if outsize is None else outsize def check_type_forward(self, in_types): n_in = in_types.size() type_check.expect(n_in == 1) x_type = in_types[0] type_check.expect( x_type.dtype.kind == 'f', x_type.ndim == 2 + self.ndim, ) if self.outs is not None: expected_dims = tuple( conv.get_conv_outsize(out, k, s, p, cover_all=self.cover_all) for (out, k, s, p) in six.moves.zip(self.outs, self.ksize, self.stride, self.pad)) type_check.expect(x_type.shape[2:] == expected_dims) def forward(self, x): self.retain_inputs(()) dims = x[0].shape[2:] ndim = self.ndim ksize = self.ksize stride = self.stride pad = self.pad if self.outs is None: self.outs = tuple( conv.get_deconv_outsize(d, k, s, p, cover_all=self.cover_all) for (d, k, s, p) in six.moves.zip(dims, ksize, stride, pad)) xp = backend.get_array_module(*x) colon = slice(None) # (:, :, None, None, ..., None) tile_index = (colon, colon) + (None,) * ndim # (1, 1, k_1, k_2, ..., k_n, 1, 1, ..., 1) tile_reps = (1, 1) + ksize + (1,) * ndim col = xp.tile(x[0][tile_index], tile_reps) if xp is numpy: col2im_nd = conv_nd.col2im_nd_cpu else: col2im_nd = conv_nd.col2im_nd_gpu y = col2im_nd(col, stride, pad, self.outs) return y, def backward(self, indexes, grad_outputs): return UnpoolingNDGrad(self).apply(grad_outputs) class UnpoolingNDGrad(function_node.FunctionNode): def __init__(self, unpoolingnd): self.ndim = unpoolingnd.ndim self.ksize = unpoolingnd.ksize self.stride = unpoolingnd.stride self.pad = unpoolingnd.pad self.outs = unpoolingnd.outs self.cover_all = unpoolingnd.cover_all def forward(self, gy): xp = backend.get_array_module(*gy) if xp is numpy: im2col_nd = conv_nd.im2col_nd_cpu else: im2col_nd = conv_nd.im2col_nd_gpu gcol = im2col_nd( gy[0], self.ksize, self.stride, self.pad, cover_all=self.cover_all) gcol_axis = tuple(six.moves.range(2, 2 + self.ndim)) gx = gcol.sum(axis=gcol_axis) return gx, def backward(self, indexes, ggx): return UnpoolingND( self.ndim, self.ksize, self.stride, self.pad, self.outs, self.cover_all).apply(ggx) def unpooling_nd(x, ksize, stride=None, pad=0, outsize=None, cover_all=True): """Inverse operation of N-dimensional spatial pooling. .. warning:: This feature is experimental. The interface can change in the future. This function acts similarly to :class:`~functions.connection.deconvolution_nd.DeconvolutionND`, but it spreads input N-dimensional array's value without any parameter instead of computing the inner products. Args: x (~chainer.Variable): Input variable. ksize (int or pair of ints): Size of pooling window :math:`(k_1, k_2, ..., k_N)`. ``ksize=k`` is equivalent to ``(k, k, ..., k)``. stride (int, pair of ints or None): Stride of pooling applications :math:`(s_1, s_2, ..., s_N)`. ``stride=s`` is equivalent to ``(s, s, ..., s)``. If ``None`` is specified, then it uses same stride as the pooling window size. pad (int or pair of ints): Spatial padding width for the input array :math:`(p_1, p_2, ..., p_N)`. ``pad=p`` is equivalent to ``(p, p, ..., p)``. outsize (None or pair of ints): Expected output size of unpooling operation :math:`(out_1, out_2, ..., out_N)`. If ``None``, the size is estimated from input size, stride and padding. cover_all (bool): If ``True``, the pooling window is assumed to cover all of the output array, eventually the output size may be smaller than that in the case ``cover_all`` is ``False``. Returns: ~chainer.Variable: Output variable. """ ndim = len(x.shape[2:]) return UnpoolingND( ndim, ksize, stride, pad, outsize, cover_all).apply((x,))[0] def unpooling_1d(x, ksize, stride=None, pad=0, outsize=None, cover_all=True): """Inverse operation of 1-dimensional spatial pooling. .. warning:: This feature is experimental. The interface can change in the future. .. note:: This function calls :func:`~chainer.functions.unpooling_nd` internally, so see the details of the behavior in the documentation of :func:`~chainer.functions.unpooling_nd`. """ if len(x.shape[2:]) != 1: raise ValueError( 'The number of dimensions under channel dimension of the input ' '\'x\' should be 1. But the actual ndim was {}.'.format( len(x.shape[2:]))) return unpooling_nd(x, ksize, stride, pad, outsize, cover_all) def unpooling_3d(x, ksize, stride=None, pad=0, outsize=None, cover_all=True): """Inverse operation of 3-dimensional spatial pooling. .. warning:: This feature is experimental. The interface can change in the future. .. note:: This function calls :func:`~chainer.functions.unpooling_nd` internally, so see the details of the behavior in the documentation of :func:`~chainer.functions.unpooling_nd`. """ if len(x.shape[2:]) != 3: raise ValueError( 'The number of dimensions under channel dimension of the input ' '\'x\' should be 3. But the actual ndim was {}.'.format( len(x.shape[2:]))) return unpooling_nd(x, ksize, stride, pad, outsize, cover_all)
6,503
33.967742
79
py
chainer
chainer-master/chainer/functions/pooling/__init__.py
0
0
0
py
chainer
chainer-master/chainer/functions/pooling/max_pooling_nd.py
import functools from operator import mul import numpy import six import chainer from chainer import backend from chainer.backends import cuda from chainer.backends import intel64 from chainer import configuration from chainer import function_node from chainer.functions.pooling import max_pooling_nd_kernel from chainer.functions.pooling import pooling_nd from chainer.utils import conv_nd import chainerx if cuda.cudnn_enabled: _cudnn_version = cuda.cuda.cudnn.getVersion() class MaxPoolingND(pooling_nd._PoolingND): """Max pooling over a set of N-dimensional planes. .. warning:: This feature is experimental. The interface can change in the future. """ def __init__(self, ndim, ksize, stride=None, pad=0, cover_all=True, return_indices=False): super(MaxPoolingND, self).__init__( ndim, ksize, stride=stride, pad=pad, cover_all=cover_all, return_indices=return_indices) def forward_chainerx(self, x): ndim = self.ndim ksize = self.ksize stride = self.stride pad = self.pad cover_all = self.cover_all # TODO(sonots): Support return_indices in ChainerX if self.return_indices: return chainer.Fallback if x[0].device.backend.name == 'cuda': # TODO(sonots): Support more ndim in ChainerX if ndim not in [2, 3]: return chainer.Fallback y = chainerx.max_pool(x[0], ksize, stride, pad, cover_all) return y, def forward_cpu(self, x): if (self.ndim == 2 and intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(x)): return self._forward_2d_ideep(x) ksize = self.ksize stride = self.stride pad = self.pad cover_all = self.cover_all in_shape = x[0].shape in_dtype = x[0].dtype col = conv_nd.im2col_nd_cpu( x[0], ksize, stride, pad, pval=-float('inf'), cover_all=cover_all) n, c = col.shape[:2] mid = (len(col.shape) - 2) // 2 + 2 ksize = col.shape[2:mid] outs = col.shape[mid:] # (n, c, k_1 * k_2 * ... * k_N, out_1, out_2, ..., out_N) col_shape = (n, c) + (functools.reduce(mul, ksize),) + outs col = col.reshape(col_shape) # We select maximum twice, since the implementation using numpy.choose # hits its bug when kh * kw >= 32. y = col.max(axis=2) self._in_shape = in_shape self._in_dtype = in_dtype self.indexes = col.argmax(axis=2) return y, def _forward_2d_ideep(self, x): assert self.ndim == 2 kh, kw = self.ksize sy, sx = self.stride ph, pw = self.pad cover_all = self.cover_all self._in_shape = x[0].shape self._in_dtype = x[0].dtype self.retain_inputs((0,)) n, c, h, w = x[0].shape y_h = conv_nd.get_conv_outsize(h, kh, sy, ph, cover_all) assert y_h > 0, 'Height in the output should be positive.' y_w = conv_nd.get_conv_outsize(w, kw, sx, pw, cover_all) assert y_w > 0, 'Width in the output should be positive.' pd = sy * (y_h - 1) + kh - h - ph pr = sx * (y_w - 1) + kw - w - pw pp = intel64.ideep.pooling2DParam( (n, c, y_h, y_w), kh, kw, sy, sx, ph, pw, pd, pr, intel64.ideep.pooling2DParam.pooling_max) y, indexes = intel64.ideep.pooling2D.Forward( intel64.ideep.array(x[0]), pp) self.indexes = indexes return y, def forward_gpu(self, x): if chainer.should_use_cudnn('>=auto') and 2 <= self.ndim <= 3: # With cuDNN v3 or greater, use cuDNN implementation for inputs # with spatial dimensions of two or more. return self.forward_cudnn(x) ndim = self.ndim ksize = self.ksize stride = self.stride pad = self.pad cover_all = self.cover_all in_shape = x[0].shape in_dtype = x[0].dtype n, c = in_shape[:2] dims = in_shape[2:] ys = tuple(conv_nd.get_conv_outsize(d, k, s, p, cover_all) for (d, k, s, p) in six.moves.zip(dims, ksize, stride, pad)) # (n, c, y_1, y_2, ..., y_N) y_shape = (n, c) + ys y = cuda.cupy.empty(y_shape, dtype=x[0].dtype) indexes = cuda.cupy.empty(y_shape, dtype=numpy.int32) in_params, out_params, operation, name = \ max_pooling_nd_kernel.MaxPoolingNDKernelForward.generate(ndim) cuda.elementwise(in_params, out_params, operation, name)( x[0].reduced_view(), *(dims + ys + ksize + stride + pad + (y, indexes))) self._in_shape = in_shape self._in_dtype = in_dtype self.indexes = indexes return y, def backward(self, indexes, gy): return MaxPoolingNDGrad(self).apply(gy) def get_cudnn_pool_mode(self): if _cudnn_version >= 6000 and configuration.config.cudnn_deterministic: return cuda.cuda.cudnn.CUDNN_POOLING_MAX_DETERMINISTIC else: return cuda.cuda.cudnn.CUDNN_POOLING_MAX class MaxPoolingNDGrad(function_node.FunctionNode): def __init__(self, func): self.func = func def forward_cpu(self, gy): func = self.func if (func.ndim == 2 and intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(gy)): return self._forward_2d_ideep(gy) ndim = func.ndim ksize = func.ksize stride = func.stride pad = func.pad in_shape = func._in_shape in_dtype = func._in_dtype indexes = func.indexes n, c = gy[0].shape[:2] outs = gy[0].shape[2:] dims = in_shape[2:] prod_outs = functools.reduce(mul, outs) prod_ksize = functools.reduce(mul, ksize) gcol = numpy.zeros(n * c * prod_outs * prod_ksize, dtype=in_dtype) indexes = ( indexes.flatten() + numpy.arange(0, indexes.size * prod_ksize, prod_ksize)) gcol[indexes] = gy[0].ravel() gcol_shape = (n, c) + outs + ksize gcol = gcol.reshape(gcol_shape) for i in six.moves.range(ndim): gcol = numpy.swapaxes(gcol, 2 + i, ndim + 2 + i) gx = conv_nd.col2im_nd_cpu(gcol, stride, pad, dims) return gx, def _forward_2d_ideep(self, gy): func = self.func # FIXME # Here we expect indexes is returned from MKL-DNN # otherwise, there are dtype mismatch for reorder (int64-->uint8) if not isinstance(func.indexes, intel64.ideep.mdarray): return self.forward_cpu(gy) kh, kw = func.ksize sy, sx = func.stride ph, pw = func.pad indexes = func.indexes in_shape = func._in_shape n, c, h, w = in_shape y_h, y_w = gy[0].shape[2:] x = func.get_retained_inputs()[0].array pd = sy * (y_h - 1) + kh - h - ph pr = sx * (y_w - 1) + kw - w - pw pp = intel64.ideep.pooling2DParam( func._in_shape, kh, kw, sy, sx, ph, pw, pd, pr, intel64.ideep.pooling2DParam.pooling_max) indexes = intel64.ideep.array(indexes) gx = intel64.ideep.pooling2D.Backward( intel64.ideep.array(x), intel64.ideep.array(gy[0]), indexes, pp) return gx, def forward_gpu(self, gy): func = self.func if func.is_cudnn_used: return func.backward_cudnn(gy) ndim = func.ndim ksize = func.ksize stride = func.stride pad = func.pad in_shape = func._in_shape in_dtype = func._in_dtype indexes = backend.from_chx(func.indexes) n, c = in_shape[:2] dims = in_shape[2:] ys = gy[0].shape[2:] gx = cuda.cupy.empty(in_shape, in_dtype) in_params, out_params, operation, name = \ max_pooling_nd_kernel.MaxPoolingNDKernelBackward.generate(ndim) cuda.elementwise(in_params, out_params, operation, name)( gy[0].reduced_view(), indexes.reduced_view(), *(dims + ys + ksize + stride + pad + (gx,))) return gx, def backward(self, indexes, ggx): return MaxPoolingNDWithIndexes(self.func).apply(ggx) class MaxPoolingNDWithIndexes(function_node.FunctionNode): def __init__(self, func): self.func = func def forward_cpu(self, x): func = self.func ndim = func.ndim ksize = func.ksize stride = func.stride pad = func.pad cover_all = func.cover_all indexes = backend.from_chx(func.indexes) col = conv_nd.im2col_nd_cpu( x[0], ksize, stride, pad, pval=-float('inf'), cover_all=cover_all) n, c = col.shape[:2] mid = (len(col.shape) - 2) // 2 + 2 ksize = col.shape[2:mid] outs = col.shape[mid:] # (n, c, k_1 * k_2 * ... * k_N, out_1, out_2, ..., out_N) ksize_total = functools.reduce(mul, ksize) col_shape = (n, c) + (ksize_total,) + outs col = col.reshape(col_shape) # (n, c, out_1, ..., out_N, k_1 * .. * k_N) col_indexes = (0, 1) + tuple(six.moves.range(3, 3 + ndim)) + (2,) col = col.transpose(col_indexes) col = col.reshape(-1, ksize_total) indexes = indexes.ravel() col = col[numpy.arange(len(indexes)), indexes] return col.reshape((n, c) + outs), def forward_gpu(self, inputs): func = self.func if func.is_cudnn_used: x = func.get_retained_inputs()[0].array return self._forward_gpu_compute_indexes_again((x, inputs[0])) ndim = func.ndim ksize = func.ksize stride = func.stride pad = func.pad cover_all = func.cover_all indexes = backend.from_chx(func.indexes) x, = inputs in_shape = x.shape in_dtype = x.dtype n, c = in_shape[:2] dims = in_shape[2:] ys = tuple(conv_nd.get_conv_outsize(d, k, s, p, cover_all) for (d, k, s, p) in six.moves.zip(dims, ksize, stride, pad)) # (n, c, y_1, y_2, ..., y_N) y_shape = (n, c) + ys y = cuda.cupy.empty(y_shape, dtype=x.dtype) cls = max_pooling_nd_kernel.MaxPoolingNDKernelForwardWithIndexes in_params, out_params, operation, name = cls.generate(ndim) cuda.elementwise(in_params, out_params, operation, name)( x.reduced_view(), *(dims + ys + ksize + stride + pad + (indexes.reduced_view(), y))) self._in_shape = in_shape self._in_dtype = in_dtype return y, def _forward_gpu_compute_indexes_again(self, inputs): func = self.func ndim = func.ndim ksize = func.ksize stride = func.stride pad = func.pad cover_all = func.cover_all x, ggx = inputs in_shape = x.shape in_dtype = x.dtype n, c = in_shape[:2] dims = in_shape[2:] ys = tuple(conv_nd.get_conv_outsize(d, k, s, p, cover_all) for (d, k, s, p) in six.moves.zip(dims, ksize, stride, pad)) # (n, c, y_1, y_2, ..., y_N) y_shape = (n, c) + ys y = cuda.cupy.empty(y_shape, dtype=x.dtype) cls = max_pooling_nd_kernel.MaxPoolingNDKernelForwardWithIndexes1 in_params, out_params, operation, name = cls.generate(ndim) cuda.elementwise(in_params, out_params, operation, name)( x.reduced_view(), *(dims + ys + ksize + stride + pad + (ggx.reduced_view(), y))) self._in_shape = in_shape self._in_dtype = in_dtype return y, def max_pooling_nd(x, ksize, stride=None, pad=0, cover_all=True, return_indices=False): """N-dimensionally spatial max pooling function. .. warning:: This feature is experimental. The interface can change in the future. This function provides a N-dimensionally generalized version of :func:`~chainer.functions.max_pooling_2d`. This acts similarly to :func:`~chainer.functions.convolution_nd`, but it computes the maximum of input spatial patch for each channel without any parameter instead of computing the inner products. Args: x (~chainer.Variable): Input variable. ksize (int or tuple of ints): Size of pooling window. ``ksize=k`` and ``ksize=(k, k, ..., k)`` are equivalent. stride (int or tuple of ints or None): Stride of pooling applications. ``stride=s`` and ``stride=(s,s, ..., s)`` are equivalent. If ``None`` is specified, then it uses same stride as the pooling window size. pad (int or tuple of ints): Spatial padding width for the input array. ``pad=p`` and ``pad=(p, p, ..., p)`` are equivalent. cover_all (bool): If ``True``, all spatial locations are pooled into some output pixels. It may make the output size larger. return_indices (bool): If ``True``, pooling indices array is returned together with the output variable. The returned indices are expected for use by :func:`chainer.functions.upsampling_nd`. Note that cuDNN will not be used for this function if ``return_indices`` is set to ``True``, as cuDNN does not return indices information. Returns: ~chainer.Variable or tuple: When ``return_indices`` is ``False`` (default), returns the output variable. When ``True``, returns the tuple of the output variable and pooling indices (:ref:`ndarray`). Pooling indices will be on the same device as the input. """ ndim = len(x.shape[2:]) func = MaxPoolingND(ndim, ksize, stride, pad, cover_all, return_indices) if return_indices: with chainer.using_config('use_cudnn', 'never'): out = func.apply((x,))[0] return out, func.indexes return func.apply((x,))[0] def max_pooling_1d(x, ksize, stride=None, pad=0, cover_all=True, return_indices=False): """1-dimensional spatial max pooling function. .. warning:: This feature is experimental. The interface can change in the future. .. note:: This function calls :func:`~chainer.functions.max_pooling_nd` internally, so see the details of the behavior in the documentation of :func:`~chainer.functions.max_pooling_nd`. """ if len(x.shape[2:]) != 1: raise ValueError( 'The number of dimensions under channel dimension of the input ' '\'x\' should be 1. But the actual ndim was {}.'.format( len(x.shape[2:]))) return max_pooling_nd(x, ksize, stride, pad, cover_all, return_indices) def max_pooling_2d(x, ksize, stride=None, pad=0, cover_all=True, return_indices=False): """Spatial max pooling function. This function acts similarly to :func:`~chainer.functions.convolution_2d`, but it computes the maximum of input spatial patch for each channel without any parameter instead of computing the inner products. Args: x (~chainer.Variable): Input variable. ksize (int or pair of ints): Size of pooling window. ``ksize=k`` and ``ksize=(k, k)`` are equivalent. stride (int or pair of ints or None): Stride of pooling applications. ``stride=s`` and ``stride=(s, s)`` are equivalent. If ``None`` is specified, then it uses same stride as the pooling window size. pad (int or pair of ints): Spatial padding width for the input array. ``pad=p`` and ``pad=(p, p)`` are equivalent. cover_all (bool): If ``True``, all spatial locations are pooled into some output pixels. It may make the output size larger. return_indices (bool): If ``True``, pooling indices array is returned together with the output variable. The returned indices are expected for use by :func:`chainer.functions.upsampling_2d`. Note that cuDNN will not be used for this function if ``return_indices`` is set to ``True``, as cuDNN does not return indices information. Returns: ~chainer.Variable or tuple: When ``return_indices`` is ``False`` (default), returns the output variable. When ``True``, returns the tuple of the output variable and pooling indices (:ref:`ndarray`). Pooling indices will be on the same device as the input. """ if len(x.shape[2:]) != 2: raise ValueError( 'The number of dimensions under channel dimension of the input ' '\'x\' should be 2. But the actual ndim was {}.'.format( len(x.shape[2:]))) return max_pooling_nd(x, ksize, stride, pad, cover_all, return_indices) def max_pooling_3d(x, ksize, stride=None, pad=0, cover_all=True, return_indices=False): """3-dimensional spatial max pooling function. .. warning:: This feature is experimental. The interface can change in the future. .. note:: This function calls :func:`~chainer.functions.max_pooling_nd` internally, so see the details of the behavior in the documentation of :func:`~chainer.functions.max_pooling_nd`. """ if len(x.shape[2:]) != 3: raise ValueError( 'The number of dimensions under channel dimension of the input ' '\'x\' should be 3. But the actual ndim was {}.'.format( len(x.shape[2:]))) return max_pooling_nd(x, ksize, stride, pad, cover_all, return_indices)
17,994
33.605769
79
py
chainer
chainer-master/chainer/functions/pooling/unpooling_2d.py
import numpy import numpy.lib.stride_tricks try: import cupy.lib.stride_tricks # NOQA except Exception: pass from chainer import backend from chainer.backends import cuda from chainer import function_node from chainer.functions.pooling import pooling_2d from chainer.utils import conv from chainer.utils import type_check class Unpooling2D(pooling_2d.Pooling2D): """Unpooling over a set of 2d planes.""" def __init__(self, ksize, stride=None, pad=0, outsize=None, cover_all=True): super(Unpooling2D, self).__init__(ksize, stride, pad, cover_all) self.outh, self.outw = (None, None) if outsize is None else outsize self._use_int_scale_forward = False def check_type_forward(self, in_types): n_in = in_types.size() type_check.expect(n_in == 1) x_type = in_types[0] type_check.expect( x_type.dtype.kind == 'f', x_type.ndim == 4, ) if self.outh is not None: expected_h = conv.get_conv_outsize( self.outh, self.kh, self.sy, self.ph, cover_all=self.cover_all) type_check.expect(x_type.shape[2] == expected_h) if self.outw is not None: expected_w = conv.get_conv_outsize( self.outw, self.kw, self.sx, self.pw, cover_all=self.cover_all) type_check.expect(x_type.shape[3] == expected_w) def _integer_scale_forward(self, x): xp = backend.get_array_module(x) b, c, h, w = x.shape bs, cs, hs, ws = x.strides if self.ph > 0 or self.pw > 0: x = x[:, :, self.ph // 2:-self.ph // 2, self.pw // 2:-self.pw // 2] y = xp.lib.stride_tricks.as_strided( x, (b, c, h - self.ph, self.kh, w - self.pw, self.kw), (bs, cs, hs, 0, ws, 0)) y = y.reshape((b, c, self.kh * (h - self.ph), self.kw * (w - self.pw))) return y, def forward(self, x): h, w = x[0].shape[2:] if self.outh is None: self.outh = conv.get_deconv_outsize( h, self.kh, self.sy, self.ph, cover_all=self.cover_all) if self.outw is None: self.outw = conv.get_deconv_outsize( w, self.kw, self.sx, self.pw, cover_all=self.cover_all) if (self.outh % (h - self.ph) == 0 and self.outw % (w - self.pw) == 0 and self.outh // (h - self.ph) == self.kh and self.outw // (w - self.pw) == self.kw and self.ph % 2 == 0 and self.pw % 2 == 0 and self.sx == self.kh and self.sy == self.kw): self._use_int_scale_forward = True return self._integer_scale_forward(x[0]) xp = backend.get_array_module(*x) col = xp.tile(x[0][:, :, None, None], (1, 1, self.kh, self.kw, 1, 1)) if xp is numpy: y = conv.col2im_cpu(col, self.sy, self.sx, self.ph, self.pw, self.outh, self.outw) else: y = conv.col2im_gpu(col, self.sy, self.sx, self.ph, self.pw, self.outh, self.outw) return y, def backward(self, indexes, grad_outputs): return Unpooling2DGrad(self).apply(grad_outputs) class Unpooling2DGrad(function_node.FunctionNode): def __init__(self, unpooling2d): self.kh = unpooling2d.kh self.kw = unpooling2d.kw self.sy = unpooling2d.sy self.sx = unpooling2d.sx self.ph = unpooling2d.ph self.pw = unpooling2d.pw self.outh = unpooling2d.outh self.outw = unpooling2d.outw self.cover_all = unpooling2d.cover_all self._use_int_scale_forward = unpooling2d._use_int_scale_forward def _integer_scale_forward(self, gy): xp = backend.get_array_module(gy) b, c, h, w = gy.shape gx = gy.reshape((b, c, h // self.kh, self.kh, w // self.kw, self.kw)) gx = xp.rollaxis(gx, 3, 5).sum((4, 5)) if self.ph > 0 or self.pw > 0: tmp = xp.zeros((b, c, h // 2 + self.ph, w // 2 + self.pw), dtype=gx.dtype) tmp[:, :, self.ph // 2:-self.ph // 2, self.pw // 2:-self.pw // 2] = gx gx = tmp return gx, def forward(self, gy): if self._use_int_scale_forward: return self._integer_scale_forward(gy[0]) if isinstance(gy[0], cuda.ndarray): gcol = conv.im2col_gpu( gy[0], self.kh, self.kw, self.sy, self.sx, self.ph, self.pw, cover_all=self.cover_all) else: gcol = conv.im2col_cpu( gy[0], self.kh, self.kw, self.sy, self.sx, self.ph, self.pw, cover_all=self.cover_all) gx = gcol.sum(axis=(2, 3)) return gx, def backward(self, indexes, ggx): return Unpooling2D( (self.kh, self.kw), (self.sy, self.sx), (self.ph, self.pw), (self.outh, self.outw), self.cover_all).apply(ggx) def unpooling_2d(x, ksize, stride=None, pad=0, outsize=None, cover_all=True): """Inverse operation of pooling for 2d array. This function acts similarly to :class:`~functions.connection.deconvolution_2d.Deconvolution2DFunction`, but it spreads input 2d array's value without any parameter instead of computing the inner products. Args: x (~chainer.Variable): Input variable. ksize (int or pair of ints): Size of pooling window. ``ksize=k`` and ``ksize=(k, k)`` are equivalent. stride (int, pair of ints or None): Stride of pooling applications. ``stride=s`` and ``stride=(s, s)`` are equivalent. If ``None`` is specified, then it uses same stride as the pooling window size. pad (int or pair of ints): Spatial padding width for the input array. ``pad=p`` and ``pad=(p, p)`` are equivalent. outsize (None or pair of ints): Expected output size (height, width) of array after the operation. If ``None``, the size (height or width) is estimated from the size of input array in first batch with :func:`~chainer.utils.conv.get_deconv_outsize`. If outsize is not ``None``, the result of outsize applied to :func:`~chainer.utils.conv.get_conv_outsize` must be equal to the shape of the 2d array in the input batch ``x``. cover_all (bool): If ``True``, the output size may be smaller than the size if ``cover_all`` is ``False``. This flag serves to align behavior to the pooling functions which can cover all input locations, see :func:`~chainer.functions.max_pooling_2d` and :func:`~chainer.functions.convolution_2d`. Returns: ~chainer.Variable: Output variable. """ return Unpooling2D(ksize, stride, pad, outsize, cover_all).apply((x,))[0]
6,940
39.121387
79
py
chainer
chainer-master/chainer/functions/theano/theano_function.py
import six from chainer import backend from chainer.backends import cuda from chainer import function from chainer.utils import type_check class TheanoFunction(function.Function): def __init__(self, forward_func, backward_func): self.forward_func = forward_func self.backward_func = backward_func def check_type_forward(self, in_types): type_check.expect(in_types.size() == len(self.forward_func.indices)) for actual_type, input_info in six.moves.zip( in_types, self.forward_func.indices): expect_type = input_info[0].variable.type # Theano cannot check shapes of variables type_check.expect( actual_type.ndim == expect_type.ndim, actual_type.dtype == expect_type.numpy_dtype, ) def forward(self, inputs): gpu = backend.get_array_module(*inputs) is cuda.cupy inputs = [cuda.to_cpu(x) for x in inputs] outputs = self.forward_func(*inputs) if gpu: # TODO(unno): We can remove redundant gpu-cpu copy using # theano.sandbox.cuda.CudaNdarray.gpudata device = cuda.get_device_from_array(inputs) outputs = [cuda.to_gpu(x, device) for x in outputs] return tuple(outputs) def backward(self, inputs, grads): gpu = backend.get_array_module(*inputs) is cuda.cupy # TODO(unno): We can remove redundant gpu-cpu copy using # theano.sandbox.cuda.basic_ops.gpu_from_host args = [cuda.to_cpu(x) for x in inputs + grads] outputs = self.backward_func(*args) assert len(outputs) == len(inputs) if gpu: # TODO(unno): We can remove redundant gpu-cpu copy using # theano.sandbox.cuda.CudaNdarray.gpudata device = cuda.get_device_from_array(inputs) outputs = [cuda.to_gpu(x, device) for x in outputs] results = [] for o, i in zip(outputs, inputs): if i.dtype.kind != 'f': o = None elif o.dtype != i.dtype: o = o.astype(i.dtype) results.append(o) return tuple(results) def theano_function(forward_func, backward_func, *inputs): return TheanoFunction(forward_func, backward_func)(*inputs)
2,311
32.507246
76
py
chainer
chainer-master/chainer/functions/theano/__init__.py
0
0
0
py
chainer
chainer-master/chainer/functions/normalization/group_normalization.py
import numpy import chainer from chainer import backend from chainer.backends import cuda from chainer import configuration from chainer import function_node from chainer.utils import type_check if cuda.cudnn_enabled: cudnn = cuda.cudnn libcudnn = cuda.cuda.cudnn class GroupNormalization(function_node.FunctionNode): def __init__(self, groups, eps=1e-5): if not isinstance(groups, int): raise TypeError('Argument: \'groups\' type must be (int).') self.groups = groups self.eps = eps self.mean = None self.inv_std = None self.dummy_gamma = None def check_type_forward(self, in_types): type_check.expect(in_types.size() == 3) x_type, gamma_type, beta_type = in_types type_check.expect( x_type.dtype.kind == 'f', x_type.ndim >= 2, gamma_type.ndim == 1, beta_type.ndim == 1, gamma_type.dtype.kind == 'f', gamma_type.dtype == beta_type.dtype, x_type.shape[1] == gamma_type.shape[0], gamma_type.shape == beta_type.shape, ) def forward(self, inputs): if inputs[0].shape[1] % self.groups != 0: raise ValueError('The number of channels {} is not divisible by ' '\'groups\' argument {}.' .format(inputs[0].shape[1], self.groups)) xp = backend.get_array_module(*inputs) if xp is cuda.cupy and chainer.should_use_cudnn('>=auto', 5000): return self.forward_cudnn(inputs) self.retain_inputs((0, 1)) x, gamma, beta = inputs interm_dtype = numpy.promote_types(x.dtype, gamma.dtype) gamma = gamma.astype(interm_dtype, copy=False) beta = beta.astype(interm_dtype, copy=False) orig_shape = x.shape batch_size, channels = orig_shape[:2] groups = self.groups reduced_shape = (batch_size * groups, -1) x = x.reshape(reduced_shape) self.mean = x.mean(axis=1, dtype=interm_dtype) x_hat = x - self.mean[:, None] var = (x_hat * x_hat).mean(axis=1) var += self.eps self.inv_std = var del var xp.sqrt(self.inv_std, out=self.inv_std) xp.reciprocal(self.inv_std, out=self.inv_std) x_hat *= self.inv_std[:, None] y = x_hat.reshape((batch_size, channels, -1)) y *= gamma[:, None] y += beta[:, None] y = y.reshape(orig_shape) return y.astype(x.dtype, copy=False), def forward_cudnn(self, inputs): if self.eps < libcudnn.CUDNN_BN_MIN_EPSILON: raise RuntimeError( 'cuDNN does not allow an eps value ' 'less than {}.'.format(libcudnn.CUDNN_BN_MIN_EPSILON)) self.retain_inputs((0, 1)) x, gamma, beta = inputs xp = cuda.cupy interm_dtype = numpy.promote_types(x.dtype, gamma.dtype) gamma = gamma.astype(interm_dtype, copy=False) beta = beta.astype(interm_dtype, copy=False) orig_shape = x.shape batch_size, channels = orig_shape[:2] groups = self.groups cudnn_shape = (1, batch_size * groups, -1, 1) x = x.reshape(cudnn_shape) with x.device: dummy_beta = xp.zeros(batch_size * groups, dtype=beta.dtype) self.dummy_gamma = xp.ones_like(dummy_beta) x_hat, self.mean, self.inv_std = \ cudnn.batch_normalization_forward_training( x, self.dummy_gamma, dummy_beta, dummy_beta, dummy_beta, None, None, self.eps, 1.0, True, libcudnn.CUDNN_BATCHNORM_SPATIAL, configuration.config.debug) y = x_hat.reshape((batch_size, channels, -1)) cuda.elementwise( 'T gamma, T beta', 'U y', 'y = y * gamma + beta', 'groupnorm_y')(gamma[:, None], beta[:, None], y) y = y.reshape(orig_shape) return y, def backward(self, indexes, grad_outputs): x, gamma = self.get_retained_inputs() gy, = grad_outputs interm_dtype = numpy.promote_types(x.dtype, gamma.dtype) gamma = chainer.functions.cast(gamma, interm_dtype) orig_shape = x.shape batch_size = orig_shape[0] groups = self.groups reduced_shape = (batch_size * groups, -1) x = x.reshape(reduced_shape) x_ = chainer.functions.cast(x, interm_dtype) x_hat, = _XHat( self.eps, self.mean, self.inv_std, self.dummy_gamma).apply((x_,)) gx_hat, ggamma, gbeta = _ScaleShiftGrad().apply(( x_hat, gamma, chainer.functions.cast(gy, interm_dtype))) gx, = _XHatGrad( self.eps, self.mean, self.inv_std, self.dummy_gamma, x_hat.array).apply( (x_, gx_hat)) gx = gx.reshape(orig_shape) return chainer.functions.cast(gx, x.dtype), ggamma, gbeta class _ScaleShiftGrad(function_node.FunctionNode): def forward(self, inputs): self.retain_inputs((0, 1, 2)) x_hat, gamma, gy = inputs batch_size, channels = gy.shape[:2] gy = gy.reshape((batch_size, channels, -1)) reduced_shape = x_hat.shape x_hat = x_hat.reshape((batch_size, channels, -1)) gx_hat = gy * gamma[:, None] gbeta = gy.sum(axis=(0, 2)) if backend.get_array_module(x_hat) is cuda.cupy: ggamma = cuda.reduce( 'T gy, T x_hat', 'T ggamma', 'gy * x_hat', 'a + b', 'ggamma = a', '0', 'groupnorm_ggamma')(gy, x_hat, axis=(0, 2)) else: ggamma = (gy * x_hat).sum(axis=(0, 2)) gx_hat = gx_hat.reshape(reduced_shape) return gx_hat, ggamma, gbeta def backward(self, indexes, grad_outputs): x_hat, gamma, gy = self.get_retained_inputs() ggx_hat, gggamma, ggbeta = grad_outputs orig_shape = gy.shape batch_size, channels = gy.shape[:2] gy = gy.reshape((batch_size, channels, -1)) reduced_shape = x_hat.shape x_hat = x_hat.reshape((batch_size, channels, -1)) ggx_hat = ggx_hat.reshape((batch_size, channels, -1)) gx_hat2 = gggamma[:, None] * gy ggamma2 = chainer.functions.sum(ggx_hat * gy, axis=(0, 2)) ggy = (ggx_hat * gamma[:, None] + gggamma[:, None] * x_hat + ggbeta[:, None]) gx_hat2 = gx_hat2.reshape(reduced_shape) ggy = ggy.reshape(orig_shape) return gx_hat2, ggamma2, ggy class _XHat(function_node.FunctionNode): def __init__(self, eps, mean, inv_std, dummy_gamma): self.eps = eps self.mean = mean self.inv_std = inv_std self.dummy_gamma = dummy_gamma def forward_cpu(self, inputs): self.retain_inputs((0,)) x, = inputs x_hat = x - self.mean[:, None] x_hat *= self.inv_std[:, None] self.retain_outputs((0,)) return x_hat, def forward_gpu(self, inputs): self.retain_inputs((0,)) x, = inputs x_hat = cuda.elementwise( 'T x, U mean, U inv_std', 'T x_hat', 'x_hat = (x - mean) * inv_std', 'groupnorm_x_hat')(x, self.mean[:, None], self.inv_std[:, None]) self.retain_outputs((0,)) return x_hat, def backward(self, indexes, grad_outputs): x, = self.get_retained_inputs() x_hat, = self.get_retained_outputs() gx_hat, = grad_outputs return _XHatGrad( self.eps, self.mean, self.inv_std, self.dummy_gamma, x_hat.array).apply((x, gx_hat)) class _XHatGrad(function_node.FunctionNode): def __init__(self, eps, mean, inv_std, dummy_gamma, x_hat): self.eps = eps self.mean = mean self.inv_std = inv_std self.dummy_gamma = dummy_gamma self.x_hat = x_hat def forward(self, inputs): xp = backend.get_array_module(*inputs) if xp is cuda.cupy and chainer.should_use_cudnn('>=auto', 5000) and \ self.dummy_gamma is not None: return self.forward_cudnn(inputs) self.retain_inputs((0, 1)) _, gx_hat = inputs x_hat = self.x_hat self.x_hat = None gx_hat_avg = gx_hat.mean(axis=1, keepdims=True) gx_hat_x_hat_avg = (gx_hat * x_hat).mean(axis=1, keepdims=True) gx_std = gx_hat - gx_hat_avg - x_hat * gx_hat_x_hat_avg gx = self.inv_std[:, None] * gx_std self.retain_outputs((0,)) return gx, def forward_cudnn(self, inputs): if self.eps < libcudnn.CUDNN_BN_MIN_EPSILON: raise RuntimeError( 'cuDNN does not allow an eps value ' 'less than {}.'.format(libcudnn.CUDNN_BN_MIN_EPSILON)) self.retain_inputs((0, 1)) x, gx_hat = inputs self.x_hat = None # `x[None, :, :, None]` is slower because it results in a different # strides and cuDNN doesn't recognize it as a contiguous array. reduced_shape = x.shape cudnn_shape = (1,) + reduced_shape + (1,) x = x.reshape(cudnn_shape) gx_hat = gx_hat.reshape(cudnn_shape) gx, _, _ = cudnn.batch_normalization_backward( x, self.dummy_gamma, gx_hat, self.mean, self.inv_std, self.eps, True, libcudnn.CUDNN_BATCHNORM_SPATIAL, configuration.config.debug) gx = gx.reshape(reduced_shape) self.retain_outputs((0,)) return gx, def backward(self, indexes, grad_outputs): F = chainer.functions x, gx_hat = self.get_retained_inputs() gx, = self.get_retained_outputs() ggx, = grad_outputs x_hat, = _XHat( self.eps, self.mean, self.inv_std, self.dummy_gamma).apply((x,)) ret = [] if 0 in indexes: # -- sketch of gx2, which is grad of x through gx # gx = inv_std * gx_std # dgx = dinv_std * gx_std + inv_std * dgx_std # # -gx2l = (ggx * dinv_std * gx_std) / dx # = sum(ggx * gx_std) * (dinv_std / dx) # = -sum(ggx * gx_std) * inv_std^2 * x_hat / N # = -inv_std * x_hat * mean(ggx * gx) # # By `gx_std = gx_hat - gx_hat_avg - x_hat * gx_hat_x_hat_avg`, # -gx_hat2r = (ggx * inv_std * dgx_std) / dx_hat # = -inv_std * (ggx * mean(gx_hat * x_hat) + # gx_hat * mean(ggx * x_hat)) gx2l_std = x_hat * F.mean(ggx * gx, axis=1, keepdims=True) gx2l, = _MulInvStd( self.eps, self.mean, self.inv_std, self.dummy_gamma).apply((x, gx2l_std)) gx_hat2r_std = ( ggx * F.mean(gx_hat * x_hat, axis=1, keepdims=True) + gx_hat * F.mean(ggx * x_hat, axis=1, keepdims=True)) gx_hat2r, = _MulInvStd( self.eps, self.mean, self.inv_std, self.dummy_gamma).apply((x, gx_hat2r_std)) gx2r, = _XHatGrad( self.eps, self.mean, self.inv_std, self.dummy_gamma, x_hat.array).apply((x, gx_hat2r)) gx2 = -(gx2l + gx2r) ret.append(gx2) if 1 in indexes: ggx_hat, = _XHatGrad( self.eps, self.mean, self.inv_std, self.dummy_gamma, x_hat.array).apply((x, ggx)) ret.append(ggx_hat) return ret class _MulInvStd(function_node.FunctionNode): def __init__(self, eps, mean, inv_std, dummy_gamma): self.eps = eps self.mean = mean self.inv_std = inv_std self.dummy_gamma = dummy_gamma def forward(self, inputs): self.retain_inputs((0,)) _, y = inputs z = self.inv_std[:, None] * y self.retain_outputs((0,)) return z, def backward(self, indexes, grad_outputs): x, = self.get_retained_inputs() z, = self.get_retained_outputs() gz, = grad_outputs x_hat, = _XHat( self.eps, self.mean, self.inv_std, self.dummy_gamma).apply((x,)) gx_std = x_hat * chainer.functions.mean(gz * z, axis=1, keepdims=True) gx, = _MulInvStd( self.eps, self.mean, self.inv_std, self.dummy_gamma).apply((x, gx_std)) gy, = _MulInvStd( self.eps, self.mean, self.inv_std, self.dummy_gamma).apply((x, gz)) return -gx, gy def group_normalization(x, groups, gamma, beta, eps=1e-5): """Group normalization function. This function implements a "group normalization" which divides the channels into groups and computes within each group the mean and variance, then normalize by these statistics, scales and shifts them. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Batch tensors. First dimension of this value must be the size of minibatch and second dimension must be the number of channels. Moreover, this value must have one or more following dimensions, such as height and width. groups (int): The number of channel groups. This value must be a divisor of the number of channels. gamma (:class:`~chainer.Variable` or :ref:`ndarray`): Scaling parameter. beta (:class:`~chainer.Variable` or :ref:`ndarray`): Shifting parameter. eps (float): Epsilon value for numerical stability of normalization. Returns: ~chainer.Variable: The output variable which has the same shape as :math:`x`. See: `Group Normalization <https://arxiv.org/abs/1803.08494>`_ .. seealso:: :class:`~chainer.links.GroupNormalization` to manage the model parameters ``gamma`` and ``beta``. """ return GroupNormalization(groups, eps).apply((x, gamma, beta))[0]
13,988
32.953883
78
py
chainer
chainer-master/chainer/functions/normalization/l2_normalization.py
import six from chainer import backend from chainer import function_node import chainer.functions from chainer import utils from chainer.utils import type_check class _SetItemZero(function_node.FunctionNode): """Write values to mask of zero-initialized array""" def __init__(self, mask): self.mask = mask def forward(self, inputs): x, = inputs xp = backend.get_array_module(x) y = xp.zeros(self.mask.shape, x.dtype) y[self.mask] = x return y, def backward(self, indices, grad_outputs): g, = grad_outputs return g[self.mask], class NormalizeL2(function_node.FunctionNode): """L2 normalization""" def __init__(self, eps=1e-5, axis=1): self.eps = eps if isinstance(axis, six.integer_types): axis = axis, self.axis = axis def check_type_forward(self, in_types): type_check.expect(in_types.size() == 1) x_type, = in_types type_check.expect(x_type.dtype.kind == 'f') def forward(self, inputs): self.retain_inputs((0,)) x, = inputs xp = backend.get_array_module(x) # Note: Passing dtype argument to numpy.sqrt() because NumPy in # Python 2 looks to return a casted value to float32 when it takes a # float16 value. norm = (xp.sqrt(xp.sum(xp.square(x), axis=self.axis, keepdims=True), dtype=x.dtype) + x.dtype.type(self.eps)) return utils.force_array(x / norm), def backward(self, indexes, grad_outputs): x, = self.get_retained_inputs() gy, = grad_outputs F = chainer.functions norm_noeps = F.sqrt(F.sum(F.square(x), axis=self.axis, keepdims=True)) norm = norm_noeps + self.eps x_gy_reduced = F.sum((x * gy), axis=self.axis, keepdims=True) # L2 normalize with eps has continuous backward. However, # the backward is not differentiable for the indices of zero vectors. # To avoid nan in double backward, do not compute outside of mask. mask = norm_noeps.array != 0 x_gy_reduced, = _SetItemZero(mask).apply(( x_gy_reduced[mask] / norm_noeps[mask],)) gx = gy * norm - x_gy_reduced * x gx = gx / norm ** 2 return gx, def normalize(x, eps=1e-5, axis=1): """Normalize input by L2 norm. This function implements L2 normalization on a sample along the given axis/axes. No reduction is done along the normalization axis. In the case when :obj:`axis=1` and :math:`\\mathbf{x}` is a matrix of dimension :math:`(N, K)`, where :math:`N` and :math:`K` denote mini-batch size and the dimension of the input vectors, this function computes an output matrix :math:`\\mathbf{y}` of dimension :math:`(N, K)` by the following equation: .. math:: \\mathbf{y}_i = {\\mathbf{x}_i \\over \\| \\mathbf{x}_i \\|_2 + \\epsilon} :obj:`eps` is used to avoid division by zero when norm of :math:`\\mathbf{x}` along the given axis is zero. The default value of :obj:`axis` is determined for backward compatibility. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): multi-dimensional output variable. The first dimension is assumed to be the mini-batch dimension. eps (float): Epsilon value for numerical stability. axis (int or tuple of ints): Axis along which to normalize. Returns: ~chainer.Variable: The output variable which has the same shape as :math:`x`. """ return NormalizeL2(eps, axis).apply((x,))[0]
3,642
30.95614
78
py
chainer
chainer-master/chainer/functions/normalization/layer_normalization.py
from chainer import backend from chainer import function_node import chainer.functions from chainer.utils import type_check class LayerNormalization(function_node.FunctionNode): """Layer normalization""" def __init__(self, eps=1e-5): self.eps = eps def check_type_forward(self, in_types): type_check.expect(in_types.size() == 3) x_type, gamma_type, beta_type = in_types type_check.expect( x_type.dtype.kind == 'f', x_type.ndim == 2, gamma_type.ndim == 1, beta_type.ndim == 1, gamma_type.dtype == x_type.dtype, beta_type.dtype == x_type.dtype, gamma_type.shape == beta_type.shape, ) def _compute(self, xp, x): # xp: numpy, cupy, or chainer.functions mu = xp.mean(x, axis=1, keepdims=True) x_mu = x - mu squ_x_mu = xp.square(x_mu) var = xp.mean(squ_x_mu, axis=1, keepdims=True) std = xp.sqrt(var + self.eps) inv_std = 1. / std x_hat = x_mu * inv_std return x_mu, var, inv_std, x_hat def forward(self, inputs): self.retain_inputs((0, 1)) xp = backend.get_array_module(*inputs) x, gamma, beta = inputs x_mu, var, inv_std, x_hat = self._compute(xp, x) scaled_x = x_hat * gamma[None, ] shifted_x = scaled_x + beta[None, ] return shifted_x, def backward(self, indexes, grad_outputs): F = chainer.functions x, gamma = self.get_retained_inputs() gy, = grad_outputs x_mu, var, inv_std, x_hat = self._compute(F, x) g_beta = F.sum(gy, axis=0) g_scaled_x = gy g_gamma = F.sum(g_scaled_x * x_hat, axis=0) g_x_hat = g_scaled_x * gamma g_inv_std = F.sum(g_x_hat * x_mu, axis=1, keepdims=True) g_x_mu_1 = g_x_hat * inv_std g_std = g_inv_std * (- 1. / (var + self.eps)) g_var = g_std * 0.5 * inv_std n_units = x.shape[1] g_squ_x_mu = g_var * (1. / n_units) g_x_mu_2 = g_squ_x_mu * 2 * x_mu g_x_1 = g_x_mu_1 + g_x_mu_2 g_mu = F.sum(g_x_1, axis=1, keepdims=True) * (- 1.) g_x_2 = g_mu * (1. / n_units) g_x = g_x_1 + g_x_2 return g_x, g_gamma, g_beta, def layer_normalization(x, gamma, beta, eps=1e-5): """Layer normalization. This function implements a "layer normalization" which normalizes the input units by statistics that are computed along the second axis, scales and shifts them. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Batch vectors. Shape of this value must be `(batch_size, unit_size)`, e.g., the output of :func:`~chainer.functions.linear`. gamma (:class:`~chainer.Variable` or :ref:`ndarray`): Scaling vectors. beta (:class:`~chainer.Variable` or :ref:`ndarray`): Shifting vectors. Returns: ~chainer.Variable: The output variable which has the same shape as :math:`x`. See: `Layer Normalization <https://arxiv.org/abs/1607.06450>`_ .. seealso:: :class:`~chainer.links.LayerNormalization` to manage the model parameters ``gamma`` and ``beta``. """ return LayerNormalization(eps).apply((x, gamma, beta))[0]
3,301
29.293578
78
py
chainer
chainer-master/chainer/functions/normalization/batch_renormalization.py
import warnings import numpy from chainer import backend from chainer.backends import cuda from chainer import configuration from chainer import function from chainer.functions.normalization import batch_normalization from chainer.utils import type_check def _xhat(x, mean, std, expander): x_mu = x - mean[expander] x_mu /= std[expander] return x_mu class BatchRenormalizationFunction(function.Function): def __init__(self, eps=2e-5, mean=None, var=None, decay=0.9, rmax=1, dmax=0, update_statistics=True): self._running_mean = mean self._running_var = var self.rmax = rmax self.dmax = dmax self.r = None self.update_statistics = update_statistics self.eps = eps self.decay = decay def _warn_accessing_property(self): warnings.warn( 'The attributes of BatchRenormalizationFunction ' 'are deprecated. ' 'Consider setting update_statistics=True to ' 'batch_renormalization to update running statistics.', DeprecationWarning) @property def running_mean(self): self._warn_accessing_property() return self._running_mean @property def running_var(self): self._warn_accessing_property() return self._running_var def check_type_forward(self, in_types): type_check.expect(in_types.size() == 3) x_type, gamma_type, beta_type = in_types M = type_check.eval(gamma_type.ndim) type_check.expect( x_type.dtype.kind == 'f', x_type.ndim >= gamma_type.ndim + 1, x_type.shape[1:1 + M] == gamma_type.shape, # TODO(tkerola): Check shape gamma_type.dtype.kind == 'f', gamma_type.dtype == beta_type.dtype, gamma_type.shape == beta_type.shape, ) def forward(self, inputs): xp = backend.get_array_module(*inputs) x, gamma, beta = inputs # Note: we must be in train mode. assert configuration.config.train head_ndim = gamma.ndim + 1 expander = (None, Ellipsis) + (None,) * (x.ndim - head_ndim) # NOTE(tommi): cuDNN is not used since it does not support # batch renormalization axis = (0,) + tuple(range(head_ndim, x.ndim)) mean = x.mean(axis=axis, dtype=gamma.dtype) var = x.var(axis=axis, dtype=gamma.dtype) self.std = xp.sqrt(var + self.eps, dtype=var.dtype) running_sigma = xp.sqrt(self._running_var + self.eps, dtype=self._running_mean.dtype) self.r = xp.clip(self.std / running_sigma, 1.0 / self.rmax, self.rmax) d = xp.clip( (mean - self._running_mean) / running_sigma, -self.dmax, self.dmax) gamma = gamma[expander] beta = beta[expander] if xp is numpy: self.x_hat = _xhat(x, mean, self.std, expander) self.x_hat_renorm = self.x_hat * self.r[expander] + d[expander] y = gamma * self.x_hat_renorm y += beta y = y.astype(dtype=x.dtype) else: self.x_hat, self.x_hat_renorm, y = cuda.elementwise( 'T x, U mean, U std, U gamma, U beta, U r, U d', 'U x_hat, U x_hat_renorm, T y', ''' x_hat = (x - mean) / std; x_hat_renorm = x_hat * r + d; y = gamma * x_hat_renorm + beta; ''', 'brn_fwd')( x, mean[expander], self.std[expander], gamma, beta, self.r[expander], d[expander]) if self.update_statistics: m = x.size // gamma[expander].size self._running_mean *= self.decay adjust = m / max(m - 1., 1.) # unbiased estimation temp_ar = xp.array(mean) temp_ar *= (1 - self.decay) self._running_mean += temp_ar del temp_ar self._running_var *= self.decay temp_ar = xp.array(var) temp_ar *= (1 - self.decay) * adjust self._running_var += temp_ar del temp_ar return y, def backward(self, inputs, grad_outputs): x, gamma, _ = inputs gy = grad_outputs[0] head_ndim = gamma.ndim + 1 expander = (None, Ellipsis) + (None,) * (x.ndim - head_ndim) m = gamma.dtype.type(x.size // gamma.size) axis = (0,) + tuple(range(head_ndim, x.ndim)) xp = backend.get_array_module(x) # Note: we must be in train mode. assert configuration.config.train # NOTE(tommi): cuDNN is not used since it does not support # batch renormalization gbeta = gy.sum(axis=axis, dtype=gamma.dtype) ggamma = (gy * self.x_hat_renorm).sum(axis=axis) gsigma_batch = (gy * self.x_hat).sum(axis=axis) if xp is numpy: scale = (self.r * gamma / self.std)[expander] gx = scale * (gy - (self.x_hat * gsigma_batch[expander] + gbeta[expander]) / m) gx = gx.astype(dtype=x.dtype) else: inv_m = numpy.float32(1) / m gx = cuda.elementwise( 'T gy, U x_hat, U gamma, U std, U gsigma_batch, U gbeta, \ U inv_m, U r', 'T gx', 'gx = (r * gamma / std) * (gy - (x_hat * gsigma_batch + gbeta) * \ inv_m)', 'brn_bwd')( gy, self.x_hat, gamma[expander], self.std[expander], gsigma_batch[expander], gbeta[expander], inv_m, self.r[expander]) return gx, ggamma, gbeta def batch_renormalization(x, gamma, beta, rmax, dmax, eps=2e-5, running_mean=None, running_var=None, decay=0.9, update_statistics=False): """Batch renormalization function. This is an extension of batch normalization, which ensures that the training and inference models generate the same outputs that depend on individual examples rather than the entire minibatch. .. note:: This function does not perform in-place update to ``running_mean`` and ``running_var`` by default, contrary to :func:`~chainer.functions.batch_normalization`. If the function is called, it will not be possible to access the updated running mean and variance statistics, because they are members of the function object, which cannot be accessed by the caller. If it is desired to update the running statistics, call the function with ``update_statistics=True`` option. .. note:: For the consistency with Batch Normalization, this function intentionally ignores some of the theoretical flaws in Algorithm 1 of the Batch Renormalization paper: - ``F.batch_renormalization`` maintains the moving average of variances :math:`\\sigma^2`, while the original paper maintains the moving average of standard deviations :math:`\\sigma`. - ``F.batch_renormalization`` applies Bessel's correction to update the moving average of variances. See: `Batch Renormalization: Towards Reducing Minibatch Dependence in Batch-Normalized Models <https://arxiv.org/abs/1702.03275>`_ .. seealso:: :class:`~chainer.links.BatchRenormalization` to manage the model parameters (``gamma``, ``beta``) and the statistics (``running_mean``, ``running_var``). """ if running_mean is None: raise TypeError('running_mean is required') if running_var is None: raise TypeError('running_var is required') return BatchRenormalizationFunction( eps, running_mean, running_var, decay, rmax, dmax, update_statistics )(x, gamma, beta) def fixed_batch_renormalization(x, gamma, beta, mean, var, eps=2e-5): warnings.warn( 'fixed_batch_renormalization is deprecated. ' 'Use fixed_batch_normalization instead.', DeprecationWarning) with configuration.using_config('train', False): return batch_normalization.fixed_batch_normalization( x, gamma, beta, mean, var, eps )
8,316
36.129464
82
py
chainer
chainer-master/chainer/functions/normalization/decorrelated_batch_normalization.py
import numpy from chainer import backend from chainer import function_node from chainer.utils import argument from chainer.utils import type_check # {numpy: True, cupy: False} _xp_supports_batch_eigh = {} # routines for batched matrices def _eigh(a, xp): if xp not in _xp_supports_batch_eigh: try: xp.linalg.eigh(xp.ones((2, 2, 2), xp.float32)) except ValueError: _xp_supports_batch_eigh[xp] = False else: _xp_supports_batch_eigh[xp] = True if _xp_supports_batch_eigh[xp]: return xp.linalg.eigh(a) ws = [] vs = [] for ai in a: w, v = xp.linalg.eigh(ai) ws.append(w) vs.append(v) return xp.stack(ws), xp.stack(vs) def _matmul(a, b, xp): if hasattr(xp, 'matmul'): # numpy.matmul is supported from version 1.10.0 return xp.matmul(a, b) else: return xp.einsum('bij,bjk->bik', a, b) def _diag(a, xp): s0, s1 = a.shape ret = xp.zeros((s0, s1, s1), a.dtype) arange_s1 = numpy.arange(s1) ret[:, arange_s1, arange_s1] = a return ret def _calc_axis_and_m(x_shape, batch_size): m = batch_size spatial_ndim = len(x_shape) - 2 spatial_axis = tuple(range(2, 2 + spatial_ndim)) for i in spatial_axis: m *= x_shape[i] return spatial_axis, m class DecorrelatedBatchNormalization(function_node.FunctionNode): def __init__(self, groups=16, eps=2e-5, mean=None, projection=None, decay=0.9): self.groups = groups self.running_mean = mean self.running_projection = projection self.eps = eps self.decay = decay self.axis = None def check_type_forward(self, in_types): type_check.expect(in_types.size() == 1) x_type = in_types[0] type_check.expect( x_type.dtype.kind == 'f', x_type.shape[1] % self.groups == 0, ) type_check.expect( x_type.ndim >= 2, ) def forward(self, inputs): self.retain_inputs(()) x = inputs[0] xp = backend.get_array_module(x) x_shape = x.shape b, c = x_shape[:2] g = self.groups C = c // g spatial_axis, m = _calc_axis_and_m(x_shape, b) # (g, C, m) x_hat = x.transpose((1, 0) + spatial_axis).reshape(g, C, m) mean = x_hat.mean(axis=2, keepdims=True) x_hat = x_hat - mean self.eps = x.dtype.type(self.eps) eps_matrix = self.eps * xp.eye(C, dtype=x.dtype) cov = _matmul( x_hat, x_hat.transpose(0, 2, 1), xp) / x.dtype.type(m) + eps_matrix # (g, C), (g, C, C) self.eigvals, self.eigvectors = _eigh(cov, xp) U = _matmul( _diag(self.eigvals ** -0.5, xp), self.eigvectors.transpose(0, 2, 1), xp) self.y_hat_pca = _matmul(U, x_hat, xp) # PCA whitening # ZCA whitening y_hat = _matmul(self.eigvectors, self.y_hat_pca, xp) y = y_hat.reshape((c, b) + x_shape[2:]).transpose( (1, 0) + spatial_axis) # Update running statistics if self.running_mean is not None: mean = mean.squeeze(axis=2) self.running_mean *= self.decay self.running_mean += (1 - self.decay) * mean if self.running_projection is not None: adjust = m / max(m - 1., 1.) # unbiased estimation self.running_projection *= self.decay projection = _matmul(self.eigvectors, U, xp) self.running_projection += (1 - self.decay) * adjust * projection return y, def backward(self, indexes, grad_outputs): gy, = grad_outputs f = DecorrelatedBatchNormalizationGrad( self.groups, self.eigvals, self.eigvectors, self.y_hat_pca) return f.apply((gy,)) class DecorrelatedBatchNormalizationGrad(function_node.FunctionNode): def __init__(self, groups, eigvals, eigvectors, y_hat_pca): self.groups = groups self.eigvals = eigvals self.eigvectors = eigvectors self.y_hat_pca = y_hat_pca def forward(self, inputs): self.retain_inputs(()) gy = inputs[0] xp = backend.get_array_module(gy) gy_shape = gy.shape b, c = gy_shape[:2] g = self.groups C = c // g spatial_axis, m = _calc_axis_and_m(gy_shape, b) arange_C = numpy.arange(C) diag_indices = slice(None), arange_C, arange_C gy_hat = gy.transpose((1, 0) + spatial_axis).reshape(g, C, m) eigvectors = self.eigvectors eigvals = self.eigvals y_hat_pca = self.y_hat_pca gy_hat_pca = _matmul(eigvectors.transpose(0, 2, 1), gy_hat, xp) f = gy_hat_pca.mean(axis=2, keepdims=True) K = eigvals[:, :, None] - eigvals[:, None, :] valid = K != 0 # to avoid nan, use eig_i != eig_j instead of i != j K[valid] = xp.reciprocal(K[valid]) V = _diag(eigvals, xp) V_sqrt = _diag(eigvals ** 0.5, xp) V_invsqrt = _diag(eigvals ** -0.5, xp) F_c = _matmul( gy_hat_pca, y_hat_pca.transpose(0, 2, 1), xp) / gy.dtype.type(m) M = xp.zeros_like(F_c) M[diag_indices] = F_c[diag_indices] mat = K.transpose(0, 2, 1) * ( _matmul(V, F_c.transpose(0, 2, 1), xp) + _matmul(_matmul(V_sqrt, F_c, xp), V_sqrt, xp) ) S = mat + mat.transpose(0, 2, 1) R = gy_hat_pca - f + _matmul( (S - M).transpose(0, 2, 1), y_hat_pca, xp) gx_hat = _matmul( _matmul(R.transpose(0, 2, 1), V_invsqrt, xp), eigvectors.transpose(0, 2, 1), xp ).transpose(0, 2, 1) gx = gx_hat.reshape((c, b) + gy_shape[2:]).transpose( (1, 0) + spatial_axis) self.retain_outputs(()) return gx, def backward(self, inputs, grad_outputs): # TODO(crcrpar): Implement this. raise NotImplementedError('Double backward is not implemented for' ' decorrelated batch normalization.') class FixedDecorrelatedBatchNormalization(function_node.FunctionNode): def __init__(self, groups): self.groups = groups def check_type_forward(self, in_types): type_check.expect(in_types.size() == 3) x_type, mean_type, var_type = in_types type_check.expect( x_type.dtype.kind == 'f', mean_type.dtype == x_type.dtype, var_type.dtype == x_type.dtype, ) type_check.expect( x_type.ndim >= 2, ) def forward(self, inputs): self.retain_inputs((0, 1, 2)) x, mean, projection = inputs xp = backend.get_array_module(x) x_shape = x.shape b, c = x_shape[:2] g = self.groups C = c // g spatial_axis, m = _calc_axis_and_m(x_shape, b) x_hat = x.transpose((1, 0) + spatial_axis).reshape(g, C, m) x_hat = x_hat - xp.expand_dims(mean, axis=2) y_hat = _matmul(projection, x_hat, xp) y = y_hat.reshape((c, b) + x_shape[2:]).transpose( (1, 0) + spatial_axis) return y, def backward(self, indexes, grad_outputs): x, mean, projection = self.get_retained_inputs() gy, = grad_outputs f = FixedDecorrelatedBatchNormalizationGrad(self.groups) return f.apply((x, mean, projection, gy)) class FixedDecorrelatedBatchNormalizationGrad(function_node.FunctionNode): def __init__(self, groups): self.groups = groups def forward(self, inputs): self.retain_inputs(()) x, mean, projection, gy = inputs xp = backend.get_array_module(x) gy_shape = gy.shape b, c = gy_shape[:2] g = self.groups C = c // g spatial_axis, m = _calc_axis_and_m(gy_shape, b) gy_hat = gy.transpose((1, 0) + spatial_axis).reshape(g, C, m) x_hat = x.transpose((1, 0) + spatial_axis).reshape(g, C, m) gy_hat_pca = _matmul(projection.transpose(0, 2, 1), gy_hat, xp) gx = gy_hat_pca.reshape((c, b) + gy_shape[2:]).transpose( (1, 0) + spatial_axis) rhs = x_hat - xp.expand_dims(mean, axis=2) gprojection = _matmul((x_hat - rhs).transpose(0, 2, 1), gy_hat, xp) gmean = -gy_hat_pca[..., 0] self.retain_outputs(()) return gx, gmean, gprojection def backward(self, inputs, grad_outputs): # TODO(crcrpar): Implement this. raise NotImplementedError('Double backward is not implemented for' ' fixed decorrelated batch normalization.') def decorrelated_batch_normalization(x, **kwargs): """decorrelated_batch_normalization(x, *, groups=16, eps=2e-5, \ running_mean=None, running_projection=None, decay=0.9) Decorrelated batch normalization function. It takes the input variable ``x`` and normalizes it using batch statistics to make the output zero-mean and decorrelated. Args: x (:class:`~chainer.Variable`): Input variable. groups (int): Number of groups to use for group whitening. eps (float): Epsilon value for numerical stability. running_mean (:ref:`ndarray`): Expected value of the mean. This is a running average of the mean over several mini-batches using the decay parameter. If ``None``, the expected mean is initialized to zero. running_projection (:ref:`ndarray`): Expected value of the project matrix. This is a running average of the projection over several mini-batches using the decay parameter. If ``None``, the expected projected is initialized to the identity matrix. decay (float): Decay rate of moving average. It is used during training. Returns: ~chainer.Variable: The output variable which has the same shape as :math:`x`. See: `Decorrelated Batch Normalization <https://arxiv.org/abs/1804.08450>`_ .. seealso:: :class:`~chainer.links.DecorrelatedBatchNormalization` """ groups, eps, running_mean, running_projection, decay = \ argument.parse_kwargs( kwargs, ('groups', 16), ('eps', 2e-5), ('running_mean', None), ('running_projection', None), ('decay', 0.9)) f = DecorrelatedBatchNormalization( groups, eps, running_mean, running_projection, decay) return f.apply((x,))[0] def fixed_decorrelated_batch_normalization(x, mean, projection, groups=16): """Decorrelated batch normalization function with fixed statistics. This is a variant of decorrelated batch normalization, where the mean and projection statistics are given by the caller as fixed variables. This is used in testing mode of the decorrelated batch normalization layer, where batch statistics cannot be used for prediction consistency. Args: x (:class:`~chainer.Variable`): Input variable. mean (:class:`~chainer.Variable` or :ref:`ndarray`): Shifting parameter of input. projection (:class:`~chainer.Variable` or :ref:`ndarray`): Projection matrix for decorrelation of input. groups (int): Number of groups to use for group whitening. Returns: ~chainer.Variable: The output variable which has the same shape as :math:`x`. .. seealso:: :func:`~chainer.functions.decorrelated_batch_normalization`, :class:`~chainer.links.DecorrelatedBatchNormalization` """ f = FixedDecorrelatedBatchNormalization(groups) return f.apply((x, mean, projection))[0]
11,689
32.591954
79
py
chainer
chainer-master/chainer/functions/normalization/local_response_normalization.py
import numpy import six from chainer.backends import cuda from chainer.backends import intel64 from chainer import function_node from chainer.utils import type_check def _cu_conv_sum(y, x, n): # Convolutional sum # TODO(beam2d): Use scan computation rdim = x.size // (x.shape[0] * x.shape[1]) cuda.elementwise( 'raw T x, int32 rdim, int32 N, int32 n_', 'raw T y', ''' int half_n = n_ / 2; int offset = i / rdim * N * rdim + i % rdim; float sum_part = 0; for (int j = 0; j < N + half_n; ++j) { if (j < N) { sum_part += x[offset + j * rdim]; } if (j >= n_) { sum_part -= x[offset + (j - n_) * rdim]; } if (j >= half_n) { y[offset + (j - half_n) * rdim] = sum_part; } } ''', 'lrn_conv_sum')(x, rdim, x.shape[1], n, y, size=x.shape[0] * rdim) class LocalResponseNormalization(function_node.FunctionNode): """Cross-channel normalization function used in AlexNet.""" _use_ideep = False def __init__(self, n=5, k=2, alpha=1e-4, beta=.75): self.n = n self.k = k self.alpha = alpha self.beta = beta self.scale = None self.indexes = None self.unit_scale = None def check_type_forward(self, in_types): type_check.expect(in_types.size() == 1) x_type, = in_types type_check.expect( x_type.dtype.kind == 'f', x_type.ndim >= 2, ) def forward_cpu(self, inputs): if (intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(inputs, (4,))): self._use_ideep = True return self.forward_ideep(inputs) x, = inputs self.retain_inputs((0,)) self.retain_outputs((0,)) half_n = self.n // 2 x2 = numpy.square(x) sum_part = x2.copy() for i in six.moves.range(1, half_n + 1): sum_part[:, i:] += x2[:, :-i] sum_part[:, :-i] += x2[:, i:] self.unit_scale = self.k + self.alpha * sum_part self.scale = self.unit_scale ** -self.beta y = x * self.scale return y, def forward_ideep(self, inputs): x, = inputs self.retain_inputs((0,)) self.retain_outputs((0,)) param = intel64.ideep.localResponseNormalizationParam( self.n, self.k, self.n * self.alpha, self.beta, intel64.ideep.localResponseNormalizationParam.lrn_across_channels) y, indexes = intel64.ideep.localResponseNormalization.Forward( intel64.ideep.array(x), param) self.indexes = indexes return y, def forward_gpu(self, inputs): x, = inputs self.retain_inputs((0,)) self.retain_outputs((0,)) self.y = cuda.cupy.square(x) # temporary self.scale = cuda.cupy.empty_like(self.y) _cu_conv_sum(self.scale, self.y, self.n) cuda.elementwise( 'T x, T k, T alpha, T beta', 'T y, T scale', '''scale = k + alpha * scale; y = x * pow(scale, -beta);''', 'lrn_fwd')(x, self.k, self.alpha, self.beta, self.y, self.scale) return self.y, def backward(self, indexes, grad_outputs): x, = self.get_retained_inputs() y, = self.get_retained_outputs() gy, = grad_outputs f = LocalResponseNormalizationGrad( self.n, self.k, self.alpha, self.beta, self._use_ideep, self.scale, self.indexes, self.unit_scale,) return f.apply((x, y, gy)) class LocalResponseNormalizationGrad(function_node.FunctionNode): def __init__(self, n, k, alpha, beta, use_ideep, scale=None, indexes=None, unit_scale=None): self.n = n self.k = k self.alpha = alpha self.beta = beta self._use_ideep = use_ideep self.scale = scale self.indexes = indexes self.unit_scale = unit_scale def forward_cpu(self, inputs): if self._use_ideep: return self._backward_ideep(inputs) x, y, gy = inputs half_n = self.n // 2 summand = y * gy / self.unit_scale sum_part = summand.copy() for i in six.moves.range(1, half_n + 1): sum_part[:, i:] += summand[:, :-i] sum_part[:, :-i] += summand[:, i:] gx = gy * self.scale - 2 * self.alpha * self.beta * x * sum_part return gx, def _backward_ideep(self, inputs): x, y, gy = inputs param = intel64.ideep.localResponseNormalizationParam( self.n, self.k, self.n * self.alpha, self.beta, intel64.ideep.localResponseNormalizationParam.lrn_across_channels ) gx = intel64.ideep.localResponseNormalization.Backward( intel64.ideep.array(x), intel64.ideep.array(gy), self.indexes, param) return gx, def forward_gpu(self, inputs): x, y, gy = inputs summand = cuda.elementwise( 'T scale, T y, T gy', 'T summand', 'summand = y * gy / scale', 'lrn_bwd_summand')(self.scale, y, gy) gx = cuda.cupy.empty_like(x) _cu_conv_sum(gx, summand, self.n) cuda.elementwise( ' T x, T gy, T scale, T beta, T coeff', 'T gx', 'gx = pow(scale, -beta) * gy - coeff * x * gx', 'lrn_bwd')(x, gy, self.scale, self.beta, 2 * self.alpha * self.beta, gx) return gx, def backward(self, indexes, grad_outputs): # No trivial way to implement double-backward for this function. raise NotImplementedError def local_response_normalization(x, n=5, k=2, alpha=1e-4, beta=.75): """Local response normalization across neighboring channels. This function implements normalization across channels. Let :math:`x` an input image with :math:`N` channels. Then, this function computes an output image :math:`y` by following formula: .. math:: y_i = {x_i \\over \\left( k + \\ \\alpha \\sum_{j=\\max{1, i - n/2}}^{\\min{N, i + n/2}} \\ x_j^2 \\right)^\\beta}. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. n (int): Normalization window width. k (float): Smoothing parameter. alpha (float): Normalizer scaling parameter. beta (float): Normalizer power parameter. Returns: ~chainer.Variable: Output variable. See: Section 3.3 of `ImageNet Classification with Deep Convolutional Neural Networks <https://www.cs.toronto.edu/~fritz/absps/imagenet.pdf>`_ """ return LocalResponseNormalization(n, k, alpha, beta).apply((x,))[0]
6,877
31.29108
79
py
chainer
chainer-master/chainer/functions/normalization/__init__.py
0
0
0
py
chainer
chainer-master/chainer/functions/normalization/batch_normalization.py
import warnings import numpy import six import chainer from chainer import backend from chainer.backends import cuda from chainer.backends import intel64 from chainer import configuration from chainer import function_node from chainer import memory_layouts from chainer.utils import argument from chainer.utils import collections_abc from chainer.utils import type_check import chainerx class _BatchNormalizationImpl: def forward(self, axis, gamma, x, x_layout, xp, expander, beta, eps, decay, running_mean, running_var): raise NotImplementedError() def backward(self, axis, gamma, gy, x, x_layout, xp, expander, mean, inv_std, eps, var, forward_data): raise NotImplementedError() class GeneralBatchNormalizationImpl(_BatchNormalizationImpl): def forward(self, axis, gamma, x, x_layout, xp, expander, beta, eps, decay, running_mean, running_var): interm_dtype = numpy.promote_types(x.dtype, gamma.dtype) gamma = gamma[expander].astype(interm_dtype, copy=False) beta = beta[expander].astype(interm_dtype, copy=False) mean, var = self.get_mean_and_var(axis, gamma, x, xp, interm_dtype) if xp is numpy: inv_std = numpy.reciprocal(numpy.sqrt( var + eps, dtype=interm_dtype)) else: inv_std = cuda.cupyx.rsqrt( var + eps, dtype=interm_dtype) y = _apply_bn_fwd(xp, x, mean[expander], inv_std[expander], gamma, beta) # Update running statistics if given if running_mean is not None: m = x.size // gamma.size adjust = m / max(m - 1., 1.) # unbiased estimation xp = backend.get_array_module( running_mean, running_var) if xp is chainerx: running_mean, running_var = backend.from_chx( (running_mean, running_var)) if xp is numpy: running_mean *= decay running_mean += (1 - decay) * mean running_var *= decay running_var += (1 - decay) * adjust * var else: # running_mean and running_var has the same type as x # while mean and var is interm_dtype which is promoted from x cuda.elementwise( 'T mean, T var, U decay, U adjust', 'U r_mean, U r_var', ''' r_mean = r_mean * decay + mean * (1 - decay); r_var = r_var * decay + var * (1 - decay) * adjust; ''', 'update_mean_var')(mean, var, decay, adjust, running_mean, running_var) if xp is chainerx: running_mean = backend.to_chx(running_mean) running_var = backend.to_chx(running_var) y_layout = x_layout return y, y_layout, running_mean, running_var, mean, var, inv_std, None def get_mean_and_var(self, axis, gamma, x, xp, interm_dtype): mean = x.mean(axis=axis, dtype=interm_dtype) var = x.var(axis=axis, dtype=interm_dtype) return mean, var def get_ggamma_and_gbeta(self, axis, gamma, gy, x_hat, xp): gbeta = gy.sum(axis=axis, dtype=gamma.dtype) ggamma = (gy * x_hat).sum(axis=axis, dtype=gamma.dtype) return gbeta, ggamma def backward(self, axis, gamma, gy, x, x_layout, xp, expander, mean, inv_std, eps, var, forward_data): interm_dtype = numpy.promote_types(x.dtype, gamma.dtype) if isinstance(gy, intel64.mdarray): # intel64.mdarray does not support dtype option in sum, so we # convert it to numpy here. gy = numpy.asarray(gy) x_hat = _x_hat(x, mean[expander], inv_std[expander]) assert x_hat.dtype == interm_dtype gbeta, ggamma = self.get_ggamma_and_gbeta(axis, gamma, gy, x_hat, xp) inv_m = gamma.dtype.type(1. / (x.size // gamma.size)) if xp is numpy: if (isinstance(gamma, intel64.mdarray) and interm_dtype != numpy.float32): # Convert to numpy to avoid an error of "mkldnn::error" gamma = numpy.asarray(gamma) gx = (gamma * inv_std)[expander] * ( gy - (x_hat * ggamma[expander] + gbeta[expander]) * inv_m) gx = gx.astype(dtype=x.dtype, copy=False) else: # x_hat and inv_std have a promoted type of x and gamma gx = cuda.elementwise( ''' T gy, X x_hat, U gamma, X inv_std, U ggamma, U gbeta, U inv_m ''', 'T gx', ''' gx = (gamma * inv_std) * ( gy - (x_hat * ggamma + gbeta) * inv_m) ''', 'bn_bwd')(gy, x_hat, gamma[expander], inv_std[expander], ggamma[expander], gbeta[expander], inv_m).astype(x.dtype, copy=False) return gx, None, ggamma, gbeta class _IDeepBatchNormalizationImpl(_BatchNormalizationImpl): def forward(self, axis, gamma, x, x_layout, xp, expander, beta, eps, decay, running_mean, running_var): expand_dim = False if x.ndim == 2: expand_dim = True x = x[:, :, None, None] y, mean, var, inv_std = ( intel64.ideep.batchNormalization.Forward( intel64.ideep.array(x.astype(gamma.dtype, copy=False)), intel64.ideep.array(gamma), intel64.ideep.array(beta), None, None, eps )) y = y.astype(x.dtype, copy=False) if expand_dim: y = numpy.squeeze(y, axis=(2, 3)) # Update running statistics if given if running_mean is not None: m = x.size // gamma.size adjust = m / max(m - 1., 1.) # Update running_mean if isinstance(running_mean, intel64.ideep.mdarray): running_mean.inplace_axpby( decay, (1 - decay), mean) else: running_mean *= decay running_mean += mean * (1 - decay) # Update running_var if isinstance(running_var, intel64.ideep.mdarray): running_var.inplace_axpby( decay, (1 - decay), var * adjust) else: running_var *= decay running_var += var * adjust * (1 - decay) return y, None, running_mean, running_var, mean, var, inv_std, None def backward(self, axis, gamma, gy, x, x_layout, xp, expander, mean, inv_std, eps, var, forward_data): expand_dim = False if x.ndim == 2: expand_dim = True x = x[:, :, None, None] gy = gy[:, :, None, None] gx, gW = intel64.ideep.batchNormalization.Backward( intel64.ideep.array(x.astype(gamma.dtype, copy=False)), intel64.ideep.array(gy.astype(gamma.dtype, copy=False)), mean, var, intel64.ideep.array(gamma), eps) ggamma, gbeta = gW[:2] if expand_dim: gx = numpy.squeeze(gx, axis=(2, 3)) gx = gx.astype(x.dtype, copy=False) ggamma = ggamma.astype(gamma.dtype, copy=False) gbeta = gbeta.astype(gamma.dtype, copy=False) return gx, None, ggamma, gbeta class _CudnnBatchNormalizationImpl(_BatchNormalizationImpl): def __init__(self, is_for_conv2d, cudnn_mode): self.is_for_conv2d = is_for_conv2d self.cudnn_mode = cudnn_mode def forward(self, axis, gamma, x, x_layout, xp, expander, beta, eps, decay, running_mean, running_var): if running_mean is not None: mean = running_mean var = running_var else: # Create dummies. mean = xp.zeros_like(gamma, dtype=x.dtype) var = xp.zeros_like(gamma, dtype=x.dtype) # mean and inv_std are used as buffers to save # intermediate results computed during forward pass. These buffers # are used to speed-up backward pass. cudnn_x_layout = cuda._get_cudnn_tensor_layout_x(x_layout) reserve_space, y, mean, inv_std = ( cudnn.batch_normalization_forward_training_ex( x, gamma, beta, mean, var, None, None, eps, decay, self.is_for_conv2d, self.cudnn_mode, chainer.is_debug(), d_layout=cudnn_x_layout)) y_layout = x_layout return ( y, y_layout, running_mean, running_var, mean, var, inv_std, reserve_space) def backward(self, axis, gamma, gy, x, x_layout, xp, expander, mean, inv_std, eps, var, forward_data): cudnn_x_layout = cuda._get_cudnn_tensor_layout_x(x_layout) gx, ggamma, gbeta = cudnn.batch_normalization_backward( x, gamma, gy, mean, inv_std, eps, self.is_for_conv2d, self.cudnn_mode, chainer.is_debug(), d_layout=cudnn_x_layout, reserve_space=forward_data) gx = gx.astype(x.dtype, copy=False) ggamma = ggamma.astype(gamma.dtype, copy=False) gbeta = gbeta.astype(gamma.dtype, copy=False) gx_layout = x_layout return gx, gx_layout, ggamma, gbeta if cuda.cudnn_enabled: cudnn = cuda.cudnn libcudnn = cuda.libcudnn _cudnn_version = cuda.cuda.cudnn.getVersion() def _compute_axis(x_ndim, gamma_ndim, axis): if axis is not None: return axis return (0,) + tuple(range(gamma_ndim + 1, x_ndim)) # Computes a complementary set of axis def _compute_key_axis(x_ndim, gamma_ndim, axis): axis = _compute_axis(x_ndim, gamma_ndim, axis) key_axis = tuple([i for i in range(x_ndim) if i not in axis]) return key_axis def _impl_selector(batch_norm_func, inputs): x, gamma, _ = inputs xp = backend.get_array_module(x) mode = _BNMode(x, gamma, batch_norm_func.key_axis) use_cudnn = mode.can_use_cudnn(xp) use_ideep = mode.can_use_ideep() if use_ideep: return _IDeepBatchNormalizationImpl() elif use_cudnn: return _CudnnBatchNormalizationImpl(mode.is_for_conv2d, mode.get_cudnn_mode()) else: return GeneralBatchNormalizationImpl() class BatchNormalization(function_node.FunctionNode): mean = None inv_std = None forward_data = None # Arbitrary data passed from forward to backward def __init__(self, eps=2e-5, mean=None, var=None, decay=0.9, axis=None, impl_selector=_impl_selector): self.running_mean = mean self.running_var = var # Note: cuDNN requires that eps be greater than or equals to # CUDNN_BN_MIN_EPSILON. Otherwise, an error will occur. # See CUDNN_BN_MIN_EPSILON value in cudnn.h to verify minimum allowable # value. self.eps = eps if chainer.should_use_cudnn('>=auto'): if eps < libcudnn.CUDNN_BN_MIN_EPSILON: raise RuntimeError( 'cuDNN does not allow an eps value ' 'less than {}.'.format(libcudnn.CUDNN_BN_MIN_EPSILON)) self.decay = decay if isinstance(axis, collections_abc.Sequence): for i in range(1, len(axis)): if axis[i - 1] >= axis[i]: msg = 'numbers in axis must be sorted in ascending order' raise RuntimeError(msg) elif isinstance(axis, six.integer_types): axis = axis, elif axis is not None: raise RuntimeError('axis must be int, tuple of int or None') self.axis = axis self._impl_selector = impl_selector def check_type_forward(self, in_types): type_check.expect(in_types.size() == 3) x_type, gamma_type, beta_type = in_types type_check.expect( x_type.dtype.kind == 'f', gamma_type.dtype.kind == 'f', gamma_type.dtype == beta_type.dtype, gamma_type.shape == beta_type.shape, ) _x_ndim = type_check.eval(x_type.ndim) _gamma_ndim = type_check.eval(gamma_type.ndim) _axis = _compute_axis(_x_ndim, _gamma_ndim, self.axis) type_check.expect( x_type.ndim >= len(_axis), ) _key_axis = _compute_key_axis(_x_ndim, _gamma_ndim, _axis) type_check.expect( gamma_type.ndim == len(_key_axis), ) for i in range(len(_key_axis)): type_check.expect( x_type.shape[_key_axis[i]] == gamma_type.shape[i], ) def check_layout_forward(self, inputs): # TODO(niboshi): Write input layout check pass def forward_chainerx(self, inputs): # TODO(niboshi): Support conditions implemented as fallback # Running statistics are required. if self.running_mean is None or self.running_var is None: return chainer.Fallback # Fall back if the running statistics are non-contiguous CUDA arrays # since they are not supported by cuDNN. # Assert that both running statistics belong to the same backend. if self.running_mean.device.backend.name == 'cuda' and not ( self.running_mean.is_contiguous and self.running_var.is_contiguous): return chainer.Fallback x, gamma, beta = inputs axis_chx = _chainerx_compute_axis(x.ndim, gamma.ndim, self.axis) if not _chainerx_is_supported(x, axis_chx): return chainer.Fallback y = chainerx.batch_norm( x, gamma, beta, self.running_mean, self.running_var, self.eps, self.decay, axis_chx) return y, def forward(self, inputs): self.retain_inputs((0, 1)) x, gamma, beta = inputs x_layout, _, _ = self.input_layouts self.output_layouts = (x_layout,) self.axis = _compute_axis(x.ndim, gamma.ndim, self.axis) self.key_axis = _compute_key_axis(x.ndim, gamma.ndim, self.axis) x_shape = memory_layouts._transpose_shape(x.shape, x_layout, None) if all(x_shape[i] == 1 for i in self.axis): if 0 in self.axis: warnings.warn( 'A batch with no more than one sample has been given' ' to F.batch_normalization. F.batch_normalization' ' will always output a zero tensor for such batches.' ' This could be caused by incorrect configuration in' ' your code (such as running evaluation while' ' chainer.config.train=True),' ' but could also happen in the last batch of training' ' if non-repeating iterator is used.', UserWarning) else: warnings.warn( 'F.batch_normalization received a batch with single' ' dimensions along all axes that are used for aggregating' ' statistics. F.batch_normalization' ' will always output a zero tensor for such batches.', UserWarning) # TODO(niboshi): Refactor calculation of expander and axis into a # function and call it just before they are used. # expander inserts singleton dimensions to gamma and beta so that they # can be broadcasted with x. expander = [None for _ in range(x.ndim)] for i in self.key_axis: expander[i] = slice(None) expander = tuple(expander) expander = memory_layouts._transpose_shape(expander, None, x_layout) self.expander = expander xp = backend.get_array_module(x) self._impl = self._impl_selector(self, inputs) raw_axis = self.axis t = chainer.memory_layouts._get_layout_transpose_axes( x.ndim, None, x_layout, True) if t is not None: raw_axis = tuple(t[i] for i in self.axis) ( y, y_layout, self.running_mean, self.running_var, self.mean, self.var, self.inv_std, self.forward_data) = ( self._impl.forward( axis=raw_axis, gamma=gamma, x=x, x_layout=x_layout, xp=xp, expander=expander, beta=beta, eps=self.eps, decay=self.decay, running_mean=self.running_mean, running_var=self.running_var)) self.output_layouts = (y_layout,) return y, def backward(self, indexes, grad_outputs): x, gamma = self.get_retained_inputs() gy, = grad_outputs if isinstance(self._impl, _IDeepBatchNormalizationImpl): assert self.var is not None var = self.var else: var = None f = BatchNormalizationGrad( self.eps, self.expander, self.axis, self.mean, var, self.inv_std, self.key_axis, self._impl, self.forward_data, ) return f.apply((x, gamma, gy)) class BatchNormalizationGrad(function_node.FunctionNode): def __init__(self, eps, expander, axis, mean, var, inv_std, key_axis, impl, forward_data): self.eps = eps self.expander = expander self.axis = axis self.mean = mean self.var = var # Only used in iDeep implementation self.inv_std = inv_std self.key_axis = key_axis self._impl = impl self.forward_data = forward_data def check_layout_forward(self, inputs): pass def forward(self, inputs): self.retain_inputs((0, 1, 2)) x, gamma, gy = inputs x_layout, _, _ = self.input_layouts self.output_layouts = (x_layout, None, None) expander = self.expander xp = backend.get_array_module(x) gx, gx_layout, ggamma, gbeta = self._impl.backward( self.axis, gamma, gy, x, x_layout, xp, expander, self.mean, self.inv_std, self.eps, self.var, self.forward_data) self.retain_inputs((0, 1, 2)) self.retain_outputs((0, 1)) self.output_layouts = (gx_layout, None, None) return gx, ggamma, gbeta def backward(self, indexes, grad_outputs): F = chainer.functions expander = self.expander x, gamma, gy = self.get_retained_inputs() gx1, ggamma1 = self.get_retained_outputs() ggx1, gggamma1, ggbeta1 = grad_outputs xp = backend.get_array_module(x) original_gamma_dtype = gamma.dtype if gamma.dtype != x.dtype: gamma = F.cast(gamma, x.dtype) ggamma1 = F.cast(ggamma1, x.dtype) gggamma1 = F.cast(gggamma1, x.dtype) ggbeta1 = F.cast(ggbeta1, x.dtype) # auxiliary values inv_m = gamma.dtype.type(1. / (x.size // gamma.size)) r = 0 if ggx1 is None else F.sum(gx1 * ggx1, axis=self.axis) coeff = gamma * self.inv_std coeff_m = coeff * inv_m x_hat = _x_hat(x, self.mean[expander], self.inv_std[expander]) # handle None in output gradients ggx1 = _zero_if_none(xp, ggx1, x.shape, x.dtype) gggamma1 = _zero_if_none(xp, gggamma1, gamma.shape, gamma.dtype) ggbeta1 = _zero_if_none(xp, ggbeta1, gamma.shape, gamma.dtype) gggamma2 = gggamma1 - coeff_m * F.sum(x_hat * ggx1, axis=self.axis) ggbeta2 = ggbeta1 - coeff_m * F.sum(ggx1, axis=self.axis) ggamma2 = r / gamma gx_hat2 = (gggamma2[expander] * gy - (coeff_m * ggamma1)[expander] * ggx1) gstd2 = -self.inv_std * (r + F.sum(x_hat * gx_hat2, axis=self.axis)) gmean2 = -self.inv_std * F.sum(gx_hat2, axis=self.axis) gx2 = self.inv_std[expander] * gx_hat2 + inv_m * ( gmean2[expander] + x_hat * gstd2[expander]) ggy2 = (gggamma2[expander] * x_hat + ggbeta2[expander] + coeff[expander] * ggx1) gx2 = chainer.functions.cast(gx2, x.dtype) ggy2 = chainer.functions.cast(ggy2, gy.dtype) ggamma2 = chainer.functions.cast(ggamma2, original_gamma_dtype) return gx2, ggamma2, ggy2 class FixedBatchNormalization(function_node.FunctionNode): inv_std = None inv_var = None def __init__(self, eps=2e-5, axis=None): # Note: cuDNN requires that eps be greater than or equals to # CUDNN_BN_MIN_EPSILON. Otherwise, an error will occur. # See CUDNN_BN_MIN_EPSILON value in cudnn.h to verify minimum allowable # value. self.eps = eps if chainer.should_use_cudnn('>=auto'): if eps < libcudnn.CUDNN_BN_MIN_EPSILON: raise RuntimeError( 'cuDNN does not allow an eps value ' 'less than {}.'.format(libcudnn.CUDNN_BN_MIN_EPSILON)) if isinstance(axis, collections_abc.Sequence): for i in range(1, len(axis)): if axis[i - 1] >= axis[i]: msg = 'numbers in axis must be sorted in ascending order' raise RuntimeError(msg) elif isinstance(axis, six.integer_types): axis = axis, elif axis is not None: raise RuntimeError('axis must be int, tuple of int or None') self.axis = axis def check_type_forward(self, in_types): type_check.expect(in_types.size() == 5) x_type, gamma_type, beta_type, mean_type, var_type = in_types type_check.expect( x_type.dtype.kind == 'f', # TODO(beam2d): Check shape gamma_type.dtype.kind == 'f', beta_type.dtype == gamma_type.dtype, mean_type.dtype == gamma_type.dtype, var_type.dtype == gamma_type.dtype, beta_type.shape == gamma_type.shape, mean_type.shape == gamma_type.shape, var_type.shape == gamma_type.shape, ) _x_ndim = type_check.eval(x_type.ndim) _gamma_ndim = type_check.eval(gamma_type.ndim) _axis = _compute_axis(_x_ndim, _gamma_ndim, self.axis) type_check.expect( x_type.ndim >= len(_axis), ) _key_axis = _compute_key_axis(_x_ndim, _gamma_ndim, _axis) type_check.expect( gamma_type.ndim == len(_key_axis), ) for i in range(len(_key_axis)): type_check.expect( x_type.shape[_key_axis[i]] == gamma_type.shape[i], ) def forward_chainerx(self, inputs): # TODO(niboshi): Support conditions implemented as fallback # TODO(niboshi): chainerx.fixed_batch_norm does not support backward if chainer.config.enable_backprop: return chainer.Fallback x, gamma, beta, mean, var = inputs axis_chx = _chainerx_compute_axis(x.ndim, gamma.ndim, self.axis) if not _chainerx_is_supported(x, axis_chx): return chainer.Fallback y = chainerx.fixed_batch_norm( x, gamma, beta, mean, var, self.eps, axis_chx) return y, def forward(self, inputs): self.retain_inputs((0, 1, 3, 4)) x, gamma, beta, mean, var = inputs xp = backend.get_array_module(x) self.axis = _compute_axis(x.ndim, gamma.ndim, self.axis) self.key_axis = _compute_key_axis(x.ndim, gamma.ndim, self.axis) # expander inserts singleton dimensions to gamma and beta so that they # can be broadcasted with x. expander = [None for _ in range(x.ndim)] for i in self.key_axis: expander[i] = slice(None) expander = tuple(expander) self.expander = expander mode = _BNMode(x, gamma, self.key_axis, inference=True) if mode.can_use_ideep(): # TODO(niboshi): Refactor iDeep part into a separate method expand_dim = False if x.ndim == 2: expand_dim = True x = x[:, :, None, None] y, = intel64.ideep.batchNormalization.Forward( intel64.ideep.array(x.astype(gamma.dtype, copy=False)), intel64.ideep.array(gamma), intel64.ideep.array(beta), intel64.ideep.array(mean), intel64.ideep.array(var), self.eps ) y = y.astype(x.dtype, copy=False) if expand_dim: y = numpy.squeeze(y, axis=(2, 3)) # lazy self.inv_var = None self.inv_std = None elif mode.can_use_cudnn(xp): y = cudnn.batch_normalization_forward_inference( x, gamma, beta, mean, var, self.eps, mode.is_for_conv2d, mode.get_cudnn_mode()) else: # Generic CPU and GPU implementation gamma = gamma[expander] beta = beta[expander] var = var + self.eps self.inv_var = xp.reciprocal(var) self.inv_std = xp.sqrt(self.inv_var, dtype=self.inv_var.dtype) y = _apply_bn_fwd(xp, x, mean[expander], self.inv_std[expander], gamma, beta) return y, def backward(self, indexes, grad_outputs): x, gamma, mean, var = self.get_retained_inputs() gy, = grad_outputs f = FixedBatchNormalizationGrad( self.eps, self.expander, self.axis, self.inv_std, self.inv_var) return f.apply((x, gamma, mean, var, gy)) class FixedBatchNormalizationGrad(function_node.FunctionNode): def __init__(self, eps, expander, axis, inv_std, inv_var): self.eps = eps self.expander = expander self.axis = axis self.inv_std = inv_std # may be None self.inv_var = inv_var # may be None def forward(self, inputs): self.retain_inputs((0, 1, 2, 4)) x, gamma, mean, var, gy = inputs expander = self.expander xp = backend.get_array_module(x) if self.inv_std is None or self.inv_var is None: self.inv_var = xp.reciprocal(var + self.eps) self.inv_std = xp.sqrt(self.inv_var, dtype=self.inv_var.dtype) self.gamma_over_std = gamma * self.inv_std x_hat = _x_hat(x, mean[expander], self.inv_std[expander]) gx = self.gamma_over_std[expander] * gy gbeta = gy.sum(axis=self.axis, dtype=gamma.dtype) ggamma = (x_hat * gy).sum(axis=self.axis) gmean = -self.gamma_over_std * gbeta gvar = - 0.5 * self.inv_var * ( gamma * ggamma).astype(var.dtype, copy=False) gx = gx.astype(dtype=x.dtype) self.retain_outputs((0, 1, 2, 3, 4)) return gx, ggamma, gbeta, gmean, gvar def backward(self, indexes, grad_outputs): F = chainer.functions x, gamma, mean, gy = self.get_retained_inputs() ggx1, gggamma1, ggbeta1, ggmean1, ggvar1 = grad_outputs gx1, ggamma1, gbeta1, gmean1, gvar1 = self.get_retained_outputs() # Handle None in output gradients. xp = backend.get_array_module(x) ggx1 = _zero_if_none(xp, ggx1, x.shape, x.dtype) gggamma1 = _zero_if_none(xp, gggamma1, gamma.shape, gamma.dtype) ggbeta1 = _zero_if_none(xp, ggbeta1, gamma.shape, gamma.dtype) ggmean1 = _zero_if_none(xp, ggmean1, mean.shape, mean.dtype) ggvar1 = _zero_if_none(xp, ggvar1, mean.shape, mean.dtype) if x.dtype != gamma.dtype: gamma = F.cast(gamma, x.dtype) ggamma1 = F.cast(ggamma1, x.dtype) gggamma1 = F.cast(gggamma1, x.dtype) gbeta1 = F.cast(gbeta1, x.dtype) ggbeta1 = F.cast(ggbeta1, x.dtype) mean = F.cast(mean, x.dtype) gmean1 = F.cast(gmean1, x.dtype) ggmean1 = F.cast(ggmean1, x.dtype) gvar1 = F.cast(gvar1, x.dtype) ggvar1 = F.cast(ggvar1, x.dtype) expander = self.expander x_hat = _x_hat(x, mean[expander], self.inv_std[expander]) tmp = -0.5 * ggvar1 gamma_over_var = gamma * self.inv_var g_gamma_over_var = tmp * ggamma1 gggamma2 = gggamma1 + tmp * gamma_over_var gx_hat = gy * gggamma2[expander] gx2 = self.inv_std[expander] * gx_hat gmean2 = -self.inv_std * F.sum(gx_hat, axis=self.axis) g_gamma_over_std = F.sum(ggx1 * gy, axis=self.axis) - ggmean1 * gbeta1 ggbeta2 = ggbeta1 - ggmean1 * self.gamma_over_std ggy2 = (gggamma2[expander] * x_hat + ggbeta2[expander] + self.gamma_over_std[expander] * ggx1) ggamma2 = (self.inv_var * g_gamma_over_var + self.inv_std * g_gamma_over_std) gvar2 = -(ggamma2 * gamma_over_var + 0.5 * self.inv_var * ( F.sum(x_hat * gx_hat, axis=self.axis) - self.gamma_over_std * g_gamma_over_std)) gx2 = chainer.functions.cast(gx2, x.dtype) ggy2 = chainer.functions.cast(ggy2, gy.dtype) return gx2, ggamma2, gmean2, gvar2, ggy2 class _BNMode(object): def __init__(self, x, gamma, key_axis, inference=False): is_gamma_1d = gamma.ndim == 1 # cuDNN only supports these tensor dimensions because they are # the most commonly used. If there is a need to support other # dimensions with cuDNN, we could consider reshaping the input # into a 2-dim array with channels as second dim and m=<product # of all dimensions except the 2nd dimension> as the first # dimension. # ideep only support when x.dtype == gamma.dtype self.is_for_conv2d = is_gamma_1d and x.ndim == 4 and key_axis[0] == 1 self.is_for_linear = is_gamma_1d and key_axis[0] == x.ndim - 1 self.cudnn_dim_ok = self.is_for_conv2d or self.is_for_linear self.cudnn_dtype_ok = self.is_for_conv2d or (x.dtype != numpy.float16) self.ideep_ok = ( x.dtype == gamma.dtype and is_gamma_1d and intel64.inputs_all_ready((x,))) self.inference = inference def get_cudnn_mode(self): assert self.cudnn_dim_ok if self.is_for_linear: return libcudnn.CUDNN_BATCHNORM_PER_ACTIVATION if (not self.inference and _cudnn_version >= 7000 and configuration.config.cudnn_fast_batch_normalization): return libcudnn.CUDNN_BATCHNORM_SPATIAL_PERSISTENT return libcudnn.CUDNN_BATCHNORM_SPATIAL def can_use_ideep(self): return self.ideep_ok and intel64.should_use_ideep('>=auto') def can_use_cudnn(self, xp): # TODO(bkvogel): Check for float16 support again in next cuDNN version. # cuDNN v5 batch normalization does not seem to support float16. return (xp is cuda.cupy and chainer.should_use_cudnn('>=auto', 5000) and self.cudnn_dim_ok and self.cudnn_dtype_ok) def _x_hat(x, mean, inv_std): x_mu = x - mean x_mu *= inv_std return x_mu def _chainerx_compute_axis(x_ndim, gamma_ndim, axis): # Returns processed axis for ChainerX. axis_chx = ( None if axis is None else axis if isinstance(axis, tuple) else (axis,)) axis_chx = _compute_axis(x_ndim, gamma_ndim, axis_chx) assert isinstance(axis_chx, tuple) return axis_chx def _chainerx_is_supported(x, axis_chx): # Checks if the input configuration is supported in ChainerX device = x.device if device.backend.name == 'cuda': # cuDNN batch norm restriction return (x.ndim, axis_chx) in [ (4, (0, 2, 3)), (5, (0, 2, 3, 4)), ] return True def _apply_bn_fwd(xp, x, mean, inv_std, gamma, beta): # NOTE: all arguments should be broadcasted to x.shape # (mean, inv_std, gamma, and beta have to already be expanded) if xp is numpy: x_hat = _x_hat(x, mean, inv_std) y = x_hat * gamma y += beta y = y.astype(x.dtype) else: y = cuda.elementwise( 'T x, U mean, U inv_std, U gamma, U beta', 'T y', 'y = gamma * (x - mean) * inv_std + beta', 'bn_fwd' )(x, mean, inv_std, gamma, beta) return y def _zero_if_none(xp, x, shape, dtype): # TODO(Tokui): Return broadcasted 0 instead of a zeroed array. if x is None: return xp.zeros(shape, dtype=dtype) return x def batch_normalization(x, gamma, beta, **kwargs): """batch_normalization(x, gamma, beta, eps=2e-5, running_mean=None, \ running_var=None, decay=0.9, axis=None) Batch normalization function. It takes the input variable ``x`` and two parameter variables ``gamma`` and ``beta``. The parameter variables must both have the same dimensionality, which is referred to as the channel shape. This channel shape corresponds to the dimensions in the input which are not averaged over. Since the first dimension of the input corresponds to the batch size, the second dimension of ``x`` will correspond to the first dimension of the channel shape, the third dimension of ``x`` will correspond to the second channel dimension (if it exists) and so on. Therefore, the dimensionality of the input must be at least one plus the number of channel dimensions. The total effective "batch size" will then be considered to be the product of all dimensions in ``x`` except for the channel dimensions. As an example, if the input is four dimensional and the parameter variables are one dimensional, then it is assumed that the first dimension of the input is the batch size, the second dimension is the channel size, and the remaining two dimensions are considered to be spatial dimensions that will be averaged over along with the batch size in the batch normalization computations. That is, the total batch size will be considered to be the product of all input dimensions except the second dimension. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. gamma (:class:`~chainer.Variable` or :ref:`ndarray`): Scaling parameter of normalized data. beta (:class:`~chainer.Variable` or :ref:`ndarray`): Shifting parameter of scaled normalized data. eps (float): Epsilon value for numerical stability. running_mean (:ref:`ndarray`): Running average of the mean. This is a running average of the mean over several mini-batches using the decay parameter. The function takes a previous running average, and updates the array in-place by the new running average. If ``None``, the running average is not computed. If this is ``None``, then ``runnng_var`` must also be ``None``. running_var (:ref:`ndarray`): Running average of the variance. This is a running average of the variance over several mini-batches using the decay parameter. The function takes a previous running average, and updates the array in-place by the new running average. If ``None``, the running average is not computed. If this is ``None``, then ``running_mean`` must also be ``None``. decay (float): Decay rate of moving average. It is used during training. axis (int, tuple of int or None): Axis over which normalization is performed. When axis is ``None``, it is determined from input dimensions. For example, if ``x.ndim`` is 4, axis becomes (0, 2, 3) and normalization is performed over 0th, 2nd and 3rd axis of input. If it is 2, axis becomes (0) and normalization is performed over 0th axis of input. When a tuple of int is given to this option, numbers in the tuple must be being sorted in ascending order. For example, (0, 2) is OK, but (2, 0) is not. See: `Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift <https://arxiv.org/abs/1502.03167>`_ .. seealso:: :class:`~chainer.links.BatchNormalization` to manage the model parameters (``gamma``, ``beta``) and the statistics (``running_mean``, ``running_var``). """ eps, running_mean, running_var, decay, axis = argument.parse_kwargs( kwargs, ('eps', 2e-5), ('running_mean', None), ('running_var', None), ('decay', 0.9), ('axis', None), train='train argument is not supported anymore. ' 'Use chainer.using_config') return BatchNormalization(eps, running_mean, running_var, decay, axis).apply((x, gamma, beta))[0] def fixed_batch_normalization(x, gamma, beta, mean, var, eps=2e-5, axis=None): """Batch normalization function with fixed statistics. This is a variant of batch normalization, where the mean and variance statistics are given by the caller as fixed variables. This is used on testing mode of the batch normalization layer, where batch statistics cannot be used for prediction consistency. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. gamma (:class:`~chainer.Variable` or :ref:`ndarray`): Scaling parameter of normalized data. beta (:class:`~chainer.Variable` or :ref:`ndarray`): Shifting parameter of scaled normalized data. mean (:class:`~chainer.Variable` or :ref:`ndarray`): Shifting parameter of input. var (:class:`~chainer.Variable` or :ref:`ndarray`): Square of scaling parameter of input. eps (float): Epsilon value for numerical stability. axis (int, tuple of int or None): Axis over which normalization is performed. When axis is ``None``, it is determined from input dimensions. For example, if ``x.ndim is 4``, axis becomes (0, 2, 3) and normalization is performed over 0th, 2nd and 3rd axis of input. If it is 2, axis becomes (0) and normalization is performed over 0th axis of input. When a tuple of int is given to this option, numbers in the tuple must be being sorted in ascending order. For example, (0, 2) is OK, but (2, 0) is not. .. seealso:: :func:`~chainer.functions.batch_normalization`, :class:`~chainer.links.BatchNormalization` """ return FixedBatchNormalization(eps, axis).apply((x, gamma, beta, mean, var))[0]
38,911
38.186304
79
py
chainer
chainer-master/chainer/functions/rnn/lstm.py
import numpy import six import chainer from chainer import backend from chainer.backends import cuda from chainer.backends import intel64 from chainer import function from chainer import function_node from chainer.utils import type_check import chainerx def _extract_gates(x): r = x.reshape((len(x), x.shape[1] // 4, 4) + x.shape[2:]) return [r[:, :, i] for i in six.moves.range(4)] def _sigmoid(x, xp=numpy): half = x.dtype.type(0.5) return xp.tanh(x * half) * half + half def _grad_sigmoid(x): return x * (1 - x) def _grad_grad_sigmoid(x): return x * (1 - x) * (1 - 2 * x) def _grad_tanh(x): return 1 - x * x def _grad_grad_tanh(x, gx): return -2 * x * gx _preamble = ''' template <typename T> __device__ T sigmoid(T x) { const T half = 0.5; return tanh(x * half) * half + half; } template <typename T> __device__ T grad_sigmoid(T y) { return y * (1 - y); } template <typename T> __device__ T grad_tanh(T y) { return 1 - y * y; } #define COMMON_ROUTINE \ T aa = tanh(a); \ T ai = sigmoid(i_); \ T af = sigmoid(f); \ T ao = sigmoid(o); ''' class LSTM(function_node.FunctionNode): """Long short-term memory unit with forget gate. It has two inputs (c, x) and two outputs (c, h), where c indicates the cell state. x must have four times channels compared to the number of units. """ def check_type_forward(self, in_types): type_check._argname(in_types, ('c', 'x')) c_type, x_type = in_types type_check.expect( c_type.dtype.kind == 'f', x_type.dtype == c_type.dtype, c_type.ndim >= 2, x_type.ndim >= 2, c_type.ndim == x_type.ndim, x_type.shape[0] <= c_type.shape[0], x_type.shape[1] == 4 * c_type.shape[1], ) for i in six.moves.range(2, type_check.eval(c_type.ndim)): type_check.expect(x_type.shape[i] == c_type.shape[i]) def forward_chainerx(self, inputs): c, x = inputs c_next, h = chainerx.lstm(c, x) return c_next, h def forward(self, inputs): self.retain_inputs((0, 1)) c_prev, x = inputs a, i, f, o = _extract_gates(x) batch = len(x) if isinstance(x, chainer.get_cpu_array_types()): if intel64.should_use_ideep('>=auto'): xp = intel64.ideep.get_array_module(x) else: xp = numpy a = xp.tanh(a) i = _sigmoid(i, xp) f = _sigmoid(f, xp) o = _sigmoid(o, xp) c_next = numpy.empty_like(c_prev) c_next[:batch] = a * i + f * c_prev[:batch] h = o * xp.tanh(c_next[:batch]) else: c_next = cuda.cupy.empty_like(c_prev) h = cuda.cupy.empty_like(c_next[:batch]) cuda.elementwise( 'T c_prev, T a, T i_, T f, T o', 'T c, T h', ''' COMMON_ROUTINE; c = aa * ai + af * c_prev; h = ao * tanh(c); ''', 'lstm_fwd', preamble=_preamble)( c_prev[:batch], a, i, f, o, c_next[:batch], h) c_next[batch:] = c_prev[batch:] self.retain_outputs((0,)) return c_next, h def backward(self, indexes, grads): grad_inputs = ( self.get_retained_inputs() + self.get_retained_outputs() + grads) return LSTMGrad()(*grad_inputs) class LSTMGrad(function.Function): def forward(self, inputs): xp = backend.get_array_module(*inputs) c_prev, x, c_next, gc, gh = inputs batch = len(x) gx = xp.empty_like(x) ga, gi, gf, go = _extract_gates(gx) # Consider the case that either gradient is not given if gc is None: gc_update = 0 gc_rest = 0 else: gc_update = gc[:batch] gc_rest = gc[batch:] if gh is None: gh = 0 a, i, f, o = _extract_gates(x) if xp is numpy: if intel64.should_use_ideep('>=auto'): xp = intel64.ideep.get_array_module(x) tanh_a = xp.tanh(a) sig_i = _sigmoid(i, xp) sig_f = _sigmoid(f, xp) sig_o = _sigmoid(o, xp) co = xp.tanh(c_next[:batch]) gc_prev = numpy.empty_like(c_prev) # multiply f later gc_prev[:batch] = gh * sig_o * _grad_tanh(co) + gc_update gc = gc_prev[:batch] ga[:] = gc * sig_i * _grad_tanh(tanh_a) gi[:] = gc * tanh_a * _grad_sigmoid(sig_i) gf[:] = gc * c_prev[:batch] * _grad_sigmoid(sig_f) go[:] = gh * co * _grad_sigmoid(sig_o) gc_prev[:batch] *= sig_f # multiply f here gc_prev[batch:] = gc_rest else: gc_prev = xp.empty_like(c_prev) cuda.elementwise( 'T c_prev, T c, T gc, T gh, T a, T i_, T f, T o', 'T gc_prev, T ga, T gi, T gf, T go', ''' COMMON_ROUTINE; T co = tanh(c); T temp = gh * ao * grad_tanh(co) + gc; ga = temp * ai * grad_tanh(aa); gi = temp * aa * grad_sigmoid(ai); gf = temp * c_prev * grad_sigmoid(af); go = gh * co * grad_sigmoid(ao); gc_prev = temp * af; ''', 'lstm_bwd', preamble=_preamble)( c_prev[:batch], c_next[:batch], gc_update, gh, a, i, f, o, gc_prev[:batch], ga, gi, gf, go) gc_prev[batch:] = gc_rest return gc_prev, gx def backward(self, inputs, grads): xp = backend.get_array_module(*inputs) c_prev, x, c, gc, gh = inputs ggc_prev, ggx = grads batch = len(x) gc_is_none = gc is None gh_is_none = gh is None ggc_prev_is_none = ggc_prev is None ggx_is_none = ggx is None if gc_is_none: gc = 0 if gh_is_none: gh = 0 if ggc_prev_is_none: ggc_prev = 0 if ggx_is_none: ggx = 0 gc_prev = xp.empty_like(c_prev) gx = xp.empty_like(x) gc_next = xp.empty_like(c) ggc = xp.empty_like(c_prev) ggh = xp.empty_like(c[:batch]) gc_prev[batch:] = 0 gc_next[batch:] = 0 ggc[batch:] = 0 if ggc_prev_is_none else ggc_prev[batch:] ggh[batch:] = 0 c_prev = c_prev[:batch] c = c[:batch] if not gc_is_none: gc = gc[:batch] if not ggc_prev_is_none: ggc_prev = ggc_prev[:batch] if not ggx_is_none: ggx = ggx[:batch] a, i, f, o = _extract_gates(x) if not ggx_is_none: gga, ggi, ggf, ggo = _extract_gates(ggx) else: gga = 0 ggi = 0 ggf = 0 ggo = 0 ga, gi, gf, go = _extract_gates(gx) lstm_grad_grad( c_prev, a, i, f, o, c, gc, gh, ggc_prev, gga, ggi, ggf, ggo, gc_prev[:batch], ga[:], gi[:], gf[:], go[:], gc_next[:batch], ggc[:batch], ggh[:batch]) if gc_is_none: ggc = None if gh_is_none: ggh = None return gc_prev, gx, gc_next, ggc, ggh @cuda.fuse() def lstm_grad_grad( c_prev, a, i, f, o, c, gc, gh, ggc_prev, gga, ggi, ggf, ggo, gc_prev, ga, gi, gf, go, gc_next, ggc, ggh): xp = backend.get_array_module(a) sig_o = _sigmoid(o, xp) gsig_o = _grad_sigmoid(sig_o) ggsig_o = _grad_grad_sigmoid(sig_o) sig_i = _sigmoid(i, xp) gsig_i = _grad_sigmoid(sig_i) ggsig_i = _grad_grad_sigmoid(sig_i) sig_f = _sigmoid(f, xp) gsig_f = _grad_sigmoid(sig_f) ggsig_f = _grad_grad_sigmoid(sig_f) tanh_a = xp.tanh(a) gtanh_a = _grad_tanh(tanh_a) ggtanh_a = _grad_grad_tanh(tanh_a, gtanh_a) tanh_c = xp.tanh(c) gtanh_c = _grad_tanh(tanh_c) ggtanh_c = _grad_grad_tanh(tanh_c, gtanh_c) gc_bar = gh * sig_o * gtanh_c + gc gc_prev[:] = ggf * gc_bar * gsig_f ga[:] = (gga * sig_i * ggtanh_a + ggi * gtanh_a * gsig_i) * gc_bar gi[:] = (gga * gtanh_a * gsig_i + ggi * tanh_a * ggsig_i) * gc_bar gf[:] = (ggc_prev * (gh * sig_o * gtanh_c + gc) * gsig_f + ggf * gc_bar * c_prev * ggsig_f) ggc[:] = (ggc_prev * sig_f + gga * sig_i * gtanh_a + ggi * tanh_a * gsig_i + ggf * c_prev * gsig_f) dgc_do = gh * gsig_o * gtanh_c go[:] = ggc * dgc_do + ggo * gh * tanh_c * ggsig_o dgc_dc = gh * sig_o * ggtanh_c gc_next[:] = ggc * dgc_dc + ggo * gh * gtanh_c * gsig_o ggh[:] = ggc * sig_o * gtanh_c + ggo * tanh_c * gsig_o return gc_prev, ga, gi, gf, go, gc_next, ggc, ggh def lstm(c_prev, x): """Long Short-Term Memory units as an activation function. This function implements LSTM units with forget gates. Let the previous cell state ``c_prev`` and the input array ``x``. First, the input array ``x`` is split into four arrays :math:`a, i, f, o` of the same shapes along the second axis. It means that ``x`` 's second axis must have 4 times the ``c_prev`` 's second axis. The split input arrays are corresponding to: - :math:`a` : sources of cell input - :math:`i` : sources of input gate - :math:`f` : sources of forget gate - :math:`o` : sources of output gate Second, it computes the updated cell state ``c`` and the outgoing signal ``h`` as: .. math:: c &= \\tanh(a) \\sigma(i) + c_{\\text{prev}} \\sigma(f), \\\\ h &= \\tanh(c) \\sigma(o), where :math:`\\sigma` is the elementwise sigmoid function. These are returned as a tuple of two variables. This function supports variable length inputs. The mini-batch size of the current input must be equal to or smaller than that of the previous one. When mini-batch size of ``x`` is smaller than that of ``c``, this function only updates ``c[0:len(x)]`` and doesn't change the rest of ``c``, ``c[len(x):]``. So, please sort input sequences in descending order of lengths before applying the function. Args: c_prev (:class:`~chainer.Variable` or :ref:`ndarray`): Variable that holds the previous cell state. The cell state should be a zero array or the output of the previous call of LSTM. x (:class:`~chainer.Variable` or :ref:`ndarray`): Variable that holds the sources of cell input, input gate, forget gate and output gate. It must have the second dimension whose size is four times of that of the cell state. Returns: tuple: Two :class:`~chainer.Variable` objects ``c`` and ``h``. ``c`` is the updated cell state. ``h`` indicates the outgoing signal. See the original paper proposing LSTM with forget gates: `Long Short-Term Memory in Recurrent Neural Networks <http://www.felixgers.de/papers/phd.pdf>`_. .. seealso:: :class:`~chainer.links.LSTM` .. admonition:: Example Assuming ``y`` is the current incoming signal, ``c`` is the previous cell state, and ``h`` is the previous outgoing signal from an ``lstm`` function. Each of ``y``, ``c`` and ``h`` has ``n_units`` channels. Most typical preparation of ``x`` is: >>> n_units = 100 >>> y = chainer.Variable(np.zeros((1, n_units), np.float32)) >>> h = chainer.Variable(np.zeros((1, n_units), np.float32)) >>> c = chainer.Variable(np.zeros((1, n_units), np.float32)) >>> model = chainer.Chain() >>> with model.init_scope(): ... model.w = L.Linear(n_units, 4 * n_units) ... model.v = L.Linear(n_units, 4 * n_units) >>> x = model.w(y) + model.v(h) >>> c, h = F.lstm(c, x) It corresponds to calculate the input array ``x``, or the input sources :math:`a, i, f, o`, from the current incoming signal ``y`` and the previous outgoing signal ``h``. Different parameters are used for different kind of input sources. .. note:: We use the naming rule below. - incoming signal The formal input of the formulation of LSTM (e.g. in NLP, word vector or output of lower RNN layer). The input of :class:`chainer.links.LSTM` is the *incoming signal*. - input array The array which is linear transformed from *incoming signal* and the previous outgoing signal. The *input array* contains four sources, the sources of cell input, input gate, forget gate and output gate. The input of :class:`chainer.functions.activation.lstm.LSTM` is the *input array*. """ return LSTM().apply((c_prev, x))
12,992
31.810606
79
py
chainer
chainer-master/chainer/functions/rnn/n_step_gru.py
import numpy import chainer from chainer import backend from chainer.backends import cuda from chainer.functions.activation import sigmoid from chainer.functions.activation import tanh from chainer.functions.array import concat from chainer.functions.array import split_axis from chainer.functions.connection import linear from chainer.functions.rnn import n_step_rnn from chainer.utils import argument from chainer import variable import chainerx if cuda.cudnn_enabled: cudnn = cuda.cudnn def _extract_apply_in_data(inputs): if not inputs: return False, () if chainerx.is_available(): has_chainerx_array = False # Unwrap arrays arrays = [] for x in inputs: if isinstance(x, variable.Variable): if x._has_chainerx_array: arrays.append(x._data[0]) has_chainerx_array = True else: arrays.append(x.array) else: # x is ndarray arrays.append(x) if not has_chainerx_array: if isinstance(x, chainerx.ndarray): has_chainerx_array = True return has_chainerx_array, tuple(arrays) else: return False, tuple([ x.array if isinstance(x, variable.Variable) else x for x in inputs]) def _combine_inputs(hx, ws, bs, xs, num_layers, directions): combined = [] combined.append(hx) for x in xs: combined.append(x) for n in range(num_layers): for direction in range(directions): idx = directions * n + direction for i in range(6): combined.append(ws[idx][i]) for i in range(6): combined.append(bs[idx][i]) return combined def _seperate_inputs(combined, num_layers, seq_length, directions): hx = combined[0] xs = combined[1: 1 + seq_length] ws = [] bs = [] index = 1 + seq_length for n in range(num_layers): ws.append(combined[index: index + 6]) bs.append(combined[index + 6: index + 12]) index += 12 if directions == 2: ws.append(combined[index: index + 6]) bs.append(combined[index + 6: index + 12]) index += 12 return hx, ws, bs, xs class NStepGRU(n_step_rnn.BaseNStepRNN): def __init__(self, n_layers, states, lengths, **kwargs): n_step_rnn.BaseNStepRNN.__init__( self, n_layers, states, lengths, rnn_dir='uni', rnn_mode='gru', **kwargs) class NStepBiGRU(n_step_rnn.BaseNStepRNN): def __init__(self, n_layers, states, lengths, **kwargs): n_step_rnn.BaseNStepRNN.__init__( self, n_layers, states, lengths, rnn_dir='bi', rnn_mode='gru', **kwargs) def n_step_gru( n_layers, dropout_ratio, hx, ws, bs, xs, **kwargs): """n_step_gru(n_layers, dropout_ratio, hx, ws, bs, xs) Stacked Uni-directional Gated Recurrent Unit function. This function calculates stacked Uni-directional GRU with sequences. This function gets an initial hidden state :math:`h_0`, an input sequence :math:`x`, weight matrices :math:`W`, and bias vectors :math:`b`. This function calculates hidden states :math:`h_t` for each time :math:`t` from input :math:`x_t`. .. math:: r_t &= \\sigma(W_0 x_t + W_3 h_{t-1} + b_0 + b_3) \\\\ z_t &= \\sigma(W_1 x_t + W_4 h_{t-1} + b_1 + b_4) \\\\ h'_t &= \\tanh(W_2 x_t + b_2 + r_t \\cdot (W_5 h_{t-1} + b_5)) \\\\ h_t &= (1 - z_t) \\cdot h'_t + z_t \\cdot h_{t-1} As the function accepts a sequence, it calculates :math:`h_t` for all :math:`t` with one call. Six weight matrices and six bias vectors are required for each layers. So, when :math:`S` layers exists, you need to prepare :math:`6S` weight matrices and :math:`6S` bias vectors. If the number of layers ``n_layers`` is greather than :math:`1`, input of ``k``-th layer is hidden state ``h_t`` of ``k-1``-th layer. Note that all input variables except first layer may have different shape from the first layer. Args: n_layers(int): Number of layers. dropout_ratio(float): Dropout ratio. hx (~chainer.Variable): Variable holding stacked hidden states. Its shape is ``(S, B, N)`` where ``S`` is number of layers and is equal to ``n_layers``, ``B`` is mini-batch size, and ``N`` is dimension of hidden units. ws (list of list of :class:`~chainer.Variable`): Weight matrices. ``ws[i]`` represents weights for i-th layer. Each ``ws[i]`` is a list containing six matrices. ``ws[i][j]`` is corresponding with ``W_j`` in the equation. Only ``ws[0][j]`` where ``0 <= j < 3`` is ``(N, I)`` shape as they are multiplied with input variables. All other matrices has ``(N, N)`` shape. bs (list of list of :class:`~chainer.Variable`): Bias vectors. ``bs[i]`` represnents biases for i-th layer. Each ``bs[i]`` is a list containing six vectors. ``bs[i][j]`` is corresponding with ``b_j`` in the equation. Shape of each matrix is ``(N,)`` where ``N`` is dimension of hidden units. xs (list of :class:`~chainer.Variable`): A list of :class:`~chainer.Variable` holding input values. Each element ``xs[t]`` holds input value for time ``t``. Its shape is ``(B_t, I)``, where ``B_t`` is mini-batch size for time ``t``, and ``I`` is size of input units. Note that this function supports variable length sequences. When sequneces has different lengths, sort sequences in descending order by length, and transpose the sorted sequence. :func:`~chainer.functions.transpose_sequence` transpose a list of :func:`~chainer.Variable` holding sequence. So ``xs`` needs to satisfy ``xs[t].shape[0] >= xs[t + 1].shape[0]``. Returns: tuple: This function returns a tuple containing two elements, ``hy`` and ``ys``. - ``hy`` is an updated hidden states whose shape is same as ``hx``. - ``ys`` is a list of :class:`~chainer.Variable` . Each element ``ys[t]`` holds hidden states of the last layer corresponding to an input ``xs[t]``. Its shape is ``(B_t, N)`` where ``B_t`` is mini-batch size for time ``t``, and ``N`` is size of hidden units. Note that ``B_t`` is the same value as ``xs[t]``. """ return n_step_gru_base(n_layers, dropout_ratio, hx, ws, bs, xs, use_bi_direction=False, **kwargs) def n_step_bigru( n_layers, dropout_ratio, hx, ws, bs, xs, **kwargs): """n_step_bigru(n_layers, dropout_ratio, hx, ws, bs, xs) Stacked Bi-directional Gated Recurrent Unit function. This function calculates stacked Bi-directional GRU with sequences. This function gets an initial hidden state :math:`h_0`, an input sequence :math:`x`, weight matrices :math:`W`, and bias vectors :math:`b`. This function calculates hidden states :math:`h_t` for each time :math:`t` from input :math:`x_t`. .. math:: r^{f}_t &= \\sigma(W^{f}_0 x_t + W^{f}_3 h_{t-1} + b^{f}_0 + b^{f}_3) \\\\ z^{f}_t &= \\sigma(W^{f}_1 x_t + W^{f}_4 h_{t-1} + b^{f}_1 + b^{f}_4) \\\\ h^{f'}_t &= \\tanh(W^{f}_2 x_t + b^{f}_2 + r^{f}_t \\cdot (W^{f}_5 h_{t-1} + b^{f}_5)) \\\\ h^{f}_t &= (1 - z^{f}_t) \\cdot h^{f'}_t + z^{f}_t \\cdot h_{t-1} \\\\ r^{b}_t &= \\sigma(W^{b}_0 x_t + W^{b}_3 h_{t-1} + b^{b}_0 + b^{b}_3) \\\\ z^{b}_t &= \\sigma(W^{b}_1 x_t + W^{b}_4 h_{t-1} + b^{b}_1 + b^{b}_4) \\\\ h^{b'}_t &= \\tanh(W^{b}_2 x_t + b^{b}_2 + r^{b}_t \\cdot (W^{b}_5 h_{t-1} + b^{b}_5)) \\\\ h^{b}_t &= (1 - z^{b}_t) \\cdot h^{b'}_t + z^{b}_t \\cdot h_{t-1} \\\\ h_t &= [h^{f}_t; h^{b}_t] \\\\ where :math:`W^{f}` is weight matrices for forward-GRU, :math:`W^{b}` is weight matrices for backward-GRU. As the function accepts a sequence, it calculates :math:`h_t` for all :math:`t` with one call. Six weight matrices and six bias vectors are required for each layers. So, when :math:`S` layers exists, you need to prepare :math:`6S` weight matrices and :math:`6S` bias vectors. If the number of layers ``n_layers`` is greather than :math:`1`, input of ``k``-th layer is hidden state ``h_t`` of ``k-1``-th layer. Note that all input variables except first layer may have different shape from the first layer. Args: n_layers(int): Number of layers. dropout_ratio(float): Dropout ratio. hx (:class:`~chainer.Variable`): Variable holding stacked hidden states. Its shape is ``(2S, B, N)`` where ``S`` is number of layers and is equal to ``n_layers``, ``B`` is mini-batch size, and ``N`` is dimension of hidden units. ws (list of list of :class:`~chainer.Variable`): Weight matrices. ``ws[i]`` represents weights for i-th layer. Each ``ws[i]`` is a list containing six matrices. ``ws[i][j]`` is corresponding with ``W_j`` in the equation. Only ``ws[0][j]`` where ``0 <= j < 3`` is ``(N, I)`` shape as they are multiplied with input variables. All other matrices has ``(N, N)`` shape. bs (list of list of :class:`~chainer.Variable`): Bias vectors. ``bs[i]`` represnents biases for i-th layer. Each ``bs[i]`` is a list containing six vectors. ``bs[i][j]`` is corresponding with ``b_j`` in the equation. Shape of each matrix is ``(N,)`` where ``N`` is dimension of hidden units. xs (list of :class:`~chainer.Variable`): A list of :class:`~chainer.Variable` holding input values. Each element ``xs[t]`` holds input value for time ``t``. Its shape is ``(B_t, I)``, where ``B_t`` is mini-batch size for time ``t``, and ``I`` is size of input units. Note that this function supports variable length sequences. When sequneces has different lengths, sort sequences in descending order by length, and transpose the sorted sequence. :func:`~chainer.functions.transpose_sequence` transpose a list of :func:`~chainer.Variable` holding sequence. So ``xs`` needs to satisfy ``xs[t].shape[0] >= xs[t + 1].shape[0]``. use_bi_direction (bool): If ``True``, this function uses Bi-direction GRU. Returns: tuple: This function returns a tuple containing three elements, ``hy`` and ``ys``. - ``hy`` is an updated hidden states whose shape is same as ``hx``. - ``ys`` is a list of :class:`~chainer.Variable` . Each element ``ys[t]`` holds hidden states of the last layer corresponding to an input ``xs[t]``. Its shape is ``(B_t, N)`` where ``B_t`` is mini-batch size for time ``t``, and ``N`` is size of hidden units. Note that ``B_t`` is the same value as ``xs[t]``. """ return n_step_gru_base(n_layers, dropout_ratio, hx, ws, bs, xs, use_bi_direction=True, **kwargs) def n_step_gru_base(n_layers, dropout_ratio, hx, ws, bs, xs, use_bi_direction, **kwargs): """n_step_gru_base(n_layers, dropout_ratio, hx, ws, bs, xs, \ use_bi_direction) Base function for Stack GRU/BiGRU functions. This function is used at :func:`chainer.functions.n_step_bigru` and :func:`chainer.functions.n_step_gru`. This function's behavior depends on argument ``use_bi_direction``. Args: n_layers(int): Number of layers. dropout_ratio(float): Dropout ratio. hx (:class:`~chainer.Variable`): Variable holding stacked hidden states. Its shape is ``(S, B, N)`` where ``S`` is number of layers and is equal to ``n_layers``, ``B`` is mini-batch size, and ``N`` is dimension of hidden units. Because of bi-direction, the first dimension length is ``2S``. ws (list of list of :class:`~chainer.Variable`): Weight matrices. ``ws[i]`` represents weights for i-th layer. Each ``ws[i]`` is a list containing six matrices. ``ws[i][j]`` is corresponding with ``W_j`` in the equation. Only ``ws[0][j]`` where ``0 <= j < 3`` is ``(N, I)`` shape as they are multiplied with input variables. All other matrices has ``(N, N)`` shape. bs (list of list of :class:`~chainer.Variable`): Bias vectors. ``bs[i]`` represnents biases for i-th layer. Each ``bs[i]`` is a list containing six vectors. ``bs[i][j]`` is corresponding with ``b_j`` in the equation. Shape of each matrix is ``(N,)`` where ``N`` is dimension of hidden units. xs (list of :class:`~chainer.Variable`): A list of :class:`~chainer.Variable` holding input values. Each element ``xs[t]`` holds input value for time ``t``. Its shape is ``(B_t, I)``, where ``B_t`` is mini-batch size for time ``t``, and ``I`` is size of input units. Note that this function supports variable length sequences. When sequneces has different lengths, sort sequences in descending order by length, and transpose the sorted sequence. :func:`~chainer.functions.transpose_sequence` transpose a list of :func:`~chainer.Variable` holding sequence. So ``xs`` needs to satisfy ``xs[t].shape[0] >= xs[t + 1].shape[0]``. activation (str): Activation function name. Please select ``tanh`` or ``relu``. use_bi_direction (bool): If ``True``, this function uses Bi-direction GRU. .. seealso:: :func:`chainer.functions.n_step_rnn` :func:`chainer.functions.n_step_birnn` """ if kwargs: argument.check_unexpected_kwargs( kwargs, train='train argument is not supported anymore. ' 'Use chainer.using_config', use_cudnn='use_cudnn argument is not supported anymore. ' 'Use chainer.using_config') argument.assert_kwargs_empty(kwargs) xp = backend.get_array_module(hx, hx.data) directions = 1 if use_bi_direction: directions = 2 combined = _combine_inputs(hx, ws, bs, xs, n_layers, directions) has_chainerx_array, combined = _extract_apply_in_data(combined) hx_chx, ws_chx, bs_chx, xs_chx = _seperate_inputs( combined, n_layers, len(xs), directions) if has_chainerx_array and xp is chainerx and dropout_ratio == 0: if use_bi_direction: hy, ys = chainerx.n_step_bigru( n_layers, hx_chx, ws_chx, bs_chx, xs_chx) else: hy, ys = chainerx.n_step_gru( n_layers, hx_chx, ws_chx, bs_chx, xs_chx) hy = variable.Variable._init_unchecked( hy, requires_grad=hy.is_backprop_required(), is_chainerx_array=True) ys = [variable.Variable._init_unchecked( y, requires_grad=y.is_backprop_required(), is_chainerx_array=True) for y in ys] return hy, ys if xp is cuda.cupy and chainer.should_use_cudnn('>=auto', 5000): lengths = [len(x) for x in xs] xs = chainer.functions.concat(xs, axis=0) with chainer.using_device(xs.device): states = cuda.get_cudnn_dropout_states() states.set_dropout_ratio(dropout_ratio) w = n_step_rnn.cudnn_rnn_weight_concat( n_layers, states, use_bi_direction, 'gru', ws, bs) if use_bi_direction: rnn = NStepBiGRU else: rnn = NStepGRU hy, ys = rnn(n_layers, states, lengths)(hx, w, xs) sections = numpy.cumsum(lengths[:-1]) ys = chainer.functions.split_axis(ys, sections, 0) return hy, ys else: hy, _, ys = n_step_rnn.n_step_rnn_impl( _gru, n_layers, dropout_ratio, hx, None, ws, bs, xs, use_bi_direction) return hy, ys def _gru(x, h, c, w, b): xw = concat.concat([w[0], w[1], w[2]], axis=0) hw = concat.concat([w[3], w[4], w[5]], axis=0) xb = concat.concat([b[0], b[1], b[2]], axis=0) hb = concat.concat([b[3], b[4], b[5]], axis=0) gru_x = linear.linear(x, xw, xb) gru_h = linear.linear(h, hw, hb) W_r_x, W_z_x, W_x = split_axis.split_axis(gru_x, 3, axis=1) U_r_h, U_z_h, U_x = split_axis.split_axis(gru_h, 3, axis=1) r = sigmoid.sigmoid(W_r_x + U_r_h) z = sigmoid.sigmoid(W_z_x + U_z_h) h_bar = tanh.tanh(W_x + r * U_x) return (1 - z) * h_bar + z * h, None
17,052
40.796569
78
py
chainer
chainer-master/chainer/functions/rnn/tree_lstm.py
import numpy import six import chainer from chainer import backend from chainer.backends import cuda from chainer import function from chainer.utils import type_check import chainerx def _extract_gates(x, n_split=5): """Extract gates by split. This is different from ``_extract_gates`` in lstm.py, which is as follows:: r = x.reshape((x.shape[0], x.shape[1] // 4, 4) + x.shape[2:]) return (r[:, :, i] for i in six.moves.range(4)) In other words, it thinly slices ``x`` and merge them, while this thickly slices ``x``. """ r = x.reshape( (x.shape[0], n_split, x.shape[1] // n_split) + x.shape[2:]) return (r[:, i, :] for i in six.moves.range(n_split)) def _sigmoid(x): half = x.dtype.type(0.5) return numpy.tanh(x * half) * half + half def _grad_sigmoid(x): return x * (1 - x) def _grad_tanh(x): return 1 - x * x _preamble = ''' template <typename T> __device__ T sigmoid(T x) { const T half = 0.5; return tanh(x * half) * half + half; } template <typename T> __device__ T grad_sigmoid(T y) { return y * (1 - y); } template <typename T> __device__ T grad_tanh(T y) { return 1 - y * y; } #define COMMON_ROUTINE \ T aa = tanh(a); \ T ai = sigmoid(i_); \ T ao = sigmoid(o); \ ''' class TreeLSTM(function.Function): """TreeLSTM unit with N forget gates. This have variable inputs ``(c1, c2, ..., cN, x)`` where ``x`` is (3 + N) times larger than each cell. Forget gates ``(f1, f2, ..., fN)`` can depend in different partitions of ``x[:, 3 * cell_units:]``. There are two outputs ``(c, h)``. """ def check_type_forward(self, in_types): type_check.expect(in_types.size() >= 2) c_types = in_types[:-1] x_type = in_types[-1] n_ary = len(c_types) type_check.expect(x_type.ndim >= 2) for i in six.moves.range(len(c_types)): type_check.expect( c_types[i].dtype.kind == 'f', x_type.dtype == c_types[i].dtype, c_types[i].ndim >= 2, c_types[i].ndim == x_type.ndim, x_type.shape[0] == c_types[i].shape[0], x_type.shape[1] == (3 + n_ary) * c_types[i].shape[1], ) for j in six.moves.range(2, type_check.eval(c_types[i].ndim)): type_check.expect(x_type.shape[i] == c_types[i].shape[j]) def forward_chainerx(self, inputs): return chainerx.tree_lstm(*inputs) def forward(self, inputs): cs, x = inputs[:-1], inputs[-1] n_ary = len(cs) gates = list(_extract_gates(x, 3 + n_ary)) a, i, o = gates[:3] fs = gates[3:] if isinstance(x, chainer.get_cpu_array_types()): self.a = numpy.tanh(a) self.i = _sigmoid(i) self.o = _sigmoid(o) self.fs = [_sigmoid(f) for f in fs] self.c = self.a * self.i + sum(f * c for f, c in zip(self.fs, cs)) h = self.o * numpy.tanh(self.c) else: preamble = _preamble + \ ' '.join('T af{} = sigmoid(f{});'.format(j, j) for j in six.moves.range(1, n_ary + 1)) cells_str = ', '.join('T c{}'.format(j) for j in six.moves.range(1, n_ary + 1)) fgates_str = ', '.join('T f{}'.format(j) for j in six.moves.range(1, n_ary + 1)) fc_calc_str = ' + '.join('af{} * c{}'.format(j, j) for j in six.moves.range(1, n_ary + 1)) self.c, h = cuda.elementwise( 'T a, T i_, T o, {}, {}'.format(cells_str, fgates_str), 'T c, T h', ''' COMMON_ROUTINE; c = aa * ai + {}; h = ao * tanh(c); '''.format(fc_calc_str), 'treelstm_fwd', preamble=preamble)( a, i, o, *(list(cs) + fs)) return self.c, h def backward(self, inputs, grad_outputs): xp = backend.get_array_module(*inputs) cs, x = inputs[:-1], inputs[-1] n_ary = len(cs) gc, gh = grad_outputs gx = xp.empty_like(x) gates = list(_extract_gates(gx, 3 + n_ary)) ga, gi, go = gates[:3] gfs = gates[3:] # Consider the case that either gradient is not given if gc is None: gc = 0 if gh is None: gh = 0 if xp is numpy: co = numpy.tanh(self.c) tmp = gh * self.o * _grad_tanh(co) + gc ga[:] = tmp * self.i * _grad_tanh(self.a) gi[:] = tmp * self.a * _grad_sigmoid(self.i) go[:] = gh * co * _grad_sigmoid(self.o) gcs = [] for j in six.moves.range(0, n_ary): gfs[j][:] = tmp * cs[j] * _grad_sigmoid(self.fs[j]) gcs.append(tmp * self.fs[j]) else: gates = list(_extract_gates(x, 3 + n_ary)) a, i, o = gates[:3] fs = gates[3:] gcs = [xp.empty_like(c) for c in cs] preamble = _preamble + \ ' '.join('T af{} = sigmoid(f{});'.format(j, j) for j in six.moves.range(1, n_ary + 1)) cells_str = ', '.join('T c{}'.format(j) for j in six.moves.range(1, n_ary + 1)) fgates_str = ', '.join('T f{}'.format(j) for j in six.moves.range(1, n_ary + 1)) g_cells_str = ', '.join('T gc{}'.format(j) for j in six.moves.range(1, n_ary + 1)) g_fgates_str = ', '.join('T gf{}'.format(j) for j in six.moves.range(1, n_ary + 1)) gf_calc_str = '\n '.join( 'gf{} = temp * c{} * grad_sigmoid(af{});'.format(j, j, j) for j in six.moves.range(1, n_ary + 1)) gc_calc_str = '\n '.join( 'gc{} = temp * af{};'.format(j, j) for j in six.moves.range(1, n_ary + 1)) cuda.elementwise( 'T c, T gc, T gh, T a, T i_, T o, ' + '{}, {}'.format(cells_str, fgates_str), 'T ga, T gi, T go, {}, {}'.format(g_cells_str, g_fgates_str), ''' COMMON_ROUTINE; T co = tanh(c); T temp = gh * ao * grad_tanh(co) + gc; ga = temp * ai * grad_tanh(aa); gi = temp * aa * grad_sigmoid(ai); go = gh * co * grad_sigmoid(ao); {} {} '''.format(gf_calc_str, gc_calc_str), 'treelstm_bwd', preamble=preamble)( self.c, gc, gh, a, i, o, *(list(cs) + fs + [ga, gi, go] + gcs + gfs)) return list(gcs) + [gx] def tree_lstm(*inputs): """TreeLSTM unit as an activation function. This function implements TreeLSTM units both for N-ary TreeLSTM and Child-Sum TreeLSTM. Let the children cell states :math:`c_{\\text{1}}, c_{\\text{2}}, \\dots, c_{\\text{N}}`, and the incoming signal :math:`x`. First, the incoming signal :math:`x` is split into (3 + N) arrays :math:`a, i, o, f_{\\text{1}}, f_{\\text{2}}, ..., f_{\\text{N}}` of the same shapes along the second axis. It means that :math:`x` 's second axis must have (3 + N) times of the length of each :math:`c_{n}`. The splitted input signals are corresponding to: - :math:`a` : sources of cell input - :math:`i` : sources of input gate - :math:`o` : sources of output gate - :math:`f_{n}` : sources of forget gate for n-th ary Second, it computes outputs as: .. math:: c &= \\tanh(a) \\text{sigmoid}(i) \\\\ & + c_{\\text{1}} \\text{sigmoid}(f_{\\text{1}}), \\\\ & + c_{\\text{2}} \\text{sigmoid}(f_{\\text{2}}), \\\\ & + ..., \\\\ & + c_{\\text{N}} \\text{sigmoid}(f_{\\text{N}}), \\\\ h &= \\tanh(c) \\text{sigmoid}(o). These are returned as a tuple of (N + 1) variables. Args: inputs (list of :class:`~chainer.Variable`): Variable arguments which include all cell vectors from child-nodes, and an input vector. Each of the cell vectors and the input vector is :class:`~chainer.Variable` or :ref:`ndarray`. The input vector must have the second dimension whose size is (N + 3) times of that of each cell, where N denotes the total number of cells. Returns: tuple: Two :class:`~chainer.Variable` objects ``c`` and ``h``. ``c`` is the updated cell state. ``h`` indicates the outgoing signal. See the papers for details: `Improved Semantic Representations From Tree-Structured Long Short-Term Memory Networks <https://www.aclweb.org/anthology/P15-1150>`_ and `A Fast Unified Model for Parsing and Sentence Understanding <https://arxiv.org/pdf/1603.06021.pdf>`_. Tai et al.'s N-Ary TreeLSTM is little extended in Bowman et al., and this link is based on the variant by Bowman et al. Specifically, eq. 10 in Tai et al. only has one :math:`W` matrix to be applied to :math:`x`, consistently for all children. On the other hand, Bowman et al.'s model has multiple matrices, each of which affects the forget gate for each child's cell individually. .. admonition:: Example Assuming ``y`` is the current input signal, ``c`` is the previous cell state, and ``h`` is the previous output signal from an :meth:`~chainer.functions.tree_lstm` function. Each of ``y``, ``c`` and ``h`` has ``n_units`` channels. Using 2-ary (binary) TreeLSTM, most typical preparation of ``x`` is: >>> model = chainer.Chain() >>> with model.init_scope(): ... model.w = L.Linear(10, 5 * 10) ... model.v1 = L.Linear(10, 5 * 10) ... model.v2 = L.Linear(10, 5 * 10) >>> y = np.random.uniform(-1, 1, (4, 10)).astype(np.float32) >>> h1 = np.random.uniform(-1, 1, (4, 10)).astype(np.float32) >>> h2 = np.random.uniform(-1, 1, (4, 10)).astype(np.float32) >>> c1 = np.random.uniform(-1, 1, (4, 10)).astype(np.float32) >>> c2 = np.random.uniform(-1, 1, (4, 10)).astype(np.float32) >>> x = model.w(y) + model.v1(h1) + model.v2(h2) >>> c, h = F.tree_lstm(c1, c2, x) It corresponds to calculate the input sources :math:`a, i, o, f_{\\text{1}}, f_{\\text{2}}` from the current input ``y`` and the children's outputs ``h1`` and ``h2``. Different parameters are used for different kind of input sources. """ return TreeLSTM()(*inputs)
10,927
36.296928
79
py
chainer
chainer-master/chainer/functions/rnn/slstm.py
import numpy import six from chainer import backend from chainer.backends import cuda from chainer.backends import intel64 from chainer import function from chainer import function_node from chainer.utils import type_check import chainerx def _extract_gates(x): r = x.reshape((x.shape[0], x.shape[1] // 4, 4) + x.shape[2:]) return (r[:, :, i] for i in six.moves.range(4)) def _sigmoid(x, xp=numpy): half = x.dtype.type(0.5) return xp.tanh(x * half) * half + half def _grad_sigmoid(x): return x * (1 - x) def _grad_grad_sigmoid(x): return x * (1 - x) * (1 - 2 * x) def _grad_tanh(x): return 1 - x * x def _grad_grad_tanh(x, gx): return -2 * x * gx _preamble = ''' template <typename T> __device__ T sigmoid(T x) { const T half = 0.5; return tanh(x * half) * half + half; } template <typename T> __device__ T grad_sigmoid(T y) { return y * (1 - y); } template <typename T> __device__ T grad_tanh(T y) { return 1 - y * y; } #define COMMON_ROUTINE \ T aa1 = tanh(a1); \ T ai1 = sigmoid(i1); \ T af1 = sigmoid(f1); \ T aa2 = tanh(a2); \ T ai2 = sigmoid(i2); \ T af2 = sigmoid(f2); \ T ao = sigmoid(o1 + o2); ''' class SLSTM(function_node.FunctionNode): """S-LSTM unit. It has four inputs (c1, c2, x1, x2) and two outputs (c, h), where c indicates the cell state. x1 and x2 must have four times channels compared to the number of units. """ def check_type_forward(self, in_types): type_check._argname(in_types, ('c_prev1', 'c_prev2', 'x1', 'x2')) c1_type, c2_type, x1_type, x2_type = in_types type_check.expect( c1_type.dtype.kind == 'f', c2_type.dtype == c1_type.dtype, x1_type.dtype == c1_type.dtype, x2_type.dtype == c1_type.dtype, c1_type.ndim >= 2, c2_type.ndim >= 2, x1_type.ndim >= 2, x2_type.ndim >= 2, c1_type.ndim == x1_type.ndim, c1_type.ndim == c2_type.ndim, c1_type.ndim == x2_type.ndim, c1_type.shape[0] == x1_type.shape[0], c1_type.shape[0] == c2_type.shape[0], c1_type.shape[0] == x2_type.shape[0], x1_type.shape[1] == 4 * c1_type.shape[1], x2_type.shape[1] == 4 * c2_type.shape[1], ) for i in range(2, type_check.eval(c1_type.ndim)): type_check.expect(x1_type.shape[i] == c1_type.shape[i]) type_check.expect(x2_type.shape[i] == c2_type.shape[i]) type_check.expect(x1_type.shape[i] == x2_type.shape[i]) def forward_chainerx(self, inputs): c_prev1, c_prev2, x1, x2 = inputs c, h = chainerx.slstm(c_prev1, c_prev2, x1, x2) return c, h def forward(self, inputs): self.retain_inputs((0, 1, 2, 3)) c_prev1, c_prev2, x1, x2 = inputs a1, i1, f1, o1 = _extract_gates(x1) a2, i2, f2, o2 = _extract_gates(x2) if isinstance(x1, numpy.ndarray): a1 = numpy.tanh(a1) i1 = _sigmoid(i1) f1 = _sigmoid(f1) a2 = numpy.tanh(a2) i2 = _sigmoid(i2) f2 = _sigmoid(f2) o = _sigmoid(o1 + o2) c = a1 * i1 + a2 * i2 + \ f1 * c_prev1 + f2 * c_prev2 h = o * numpy.tanh(c) else: c, h = cuda.elementwise( '''T c_prev1, T a1, T i1, T f1, T o1, T c_prev2, T a2, T i2, T f2, T o2''', 'T c, T h', ''' COMMON_ROUTINE; c = aa1 * ai1 + af1 * c_prev1 + aa2 * ai2 + af2 * c_prev2; h = ao * tanh(c); ''', 'slstm_fwd', preamble=_preamble)( c_prev1, a1, i1, f1, o1, c_prev2, a2, i2, f2, o2) self.retain_outputs((0,)) return c, h def backward(self, indexes, grads): grad_inputs = ( self.get_retained_inputs() + self.get_retained_outputs() + grads) return SLSTMGrad()(*grad_inputs) class SLSTMGrad(function.Function): def forward(self, inputs): xp = backend.get_array_module(*inputs) c_prev1, c_prev2, x1, x2, c_next, gc, gh = inputs gx1 = xp.empty_like(x1) gx2 = xp.empty_like(x2) ga1, gi1, gf1, go1 = _extract_gates(gx1) ga2, gi2, gf2, go2 = _extract_gates(gx2) if gc is None: gc = 0 if gh is None: gh = 0 a1, i1, f1, o1 = _extract_gates(x1) a2, i2, f2, o2 = _extract_gates(x2) if xp is numpy: if intel64.should_use_ideep('>=auto'): xp = intel64.ideep.get_array_module(x1) tanh_a1 = xp.tanh(a1) sig_i1 = _sigmoid(i1, xp) sig_f1 = _sigmoid(f1, xp) tanh_a2 = xp.tanh(a2) sig_i2 = _sigmoid(i2, xp) sig_f2 = _sigmoid(f2, xp) sig_o = _sigmoid(o1 + o2, xp) co = xp.tanh(c_next) # multiply f later gc_prev = gh * sig_o * _grad_tanh(co) + gc ga1[:] = gc_prev * sig_i1 * _grad_tanh(tanh_a1) gi1[:] = gc_prev * tanh_a1 * _grad_sigmoid(sig_i1) gf1[:] = gc_prev * c_prev1 * _grad_sigmoid(sig_f1) go1[:] = gh * co * _grad_sigmoid(sig_o) ga2[:] = gc_prev * sig_i2 * _grad_tanh(tanh_a2) gi2[:] = gc_prev * tanh_a2 * _grad_sigmoid(sig_i2) gf2[:] = gc_prev * c_prev2 * _grad_sigmoid(sig_f2) go2[:] = gh * co * _grad_sigmoid(sig_o) # multiply f here gc_prev1 = gc_prev * sig_f1 gc_prev2 = gc_prev * sig_f2 else: a1, i1, f1, o1 = _extract_gates(x1) a2, i2, f2, o2 = _extract_gates(x2) gc_prev1 = xp.empty_like(c_prev1) gc_prev2 = xp.empty_like(c_prev2) cuda.elementwise( '''T c_prev1, T a1, T i1, T f1, T o1, T c_prev2, T a2, T i2, T f2, T o2, T c, T gc, T gh''', '''T gc_prev1, T ga1, T gi1, T gf1, T go1, T gc_prev2, T ga2, T gi2, T gf2, T go2''', ''' COMMON_ROUTINE; T co = tanh(c); T temp = gh * ao * grad_tanh(co) + gc; ga1 = temp * ai1 * grad_tanh(aa1); gi1 = temp * aa1 * grad_sigmoid(ai1); gf1 = temp * c_prev1 * grad_sigmoid(af1); go1 = gh * co * grad_sigmoid(ao); gc_prev1 = temp * af1; ga2 = temp * ai2 * grad_tanh(aa2); gi2 = temp * aa2 * grad_sigmoid(ai2); gf2 = temp * c_prev2 * grad_sigmoid(af2); go2 = gh * co * grad_sigmoid(ao); gc_prev2 = temp * af2; ''', 'lstm_bwd', preamble=_preamble)( c_prev1, a1, i1, f1, o1, c_prev2, a2, i2, f2, o2, c_next, gc, gh, gc_prev1, ga1, gi1, gf1, go1, gc_prev2, ga2, gi2, gf2, go2) return gc_prev1, gc_prev2, gx1, gx2 def backward(self, inputs, grads): xp = backend.get_array_module(*inputs) c_prev1, c_prev2, x1, x2, c, gc, gh = inputs ggc_prev1, ggc_prev2, ggx1, ggx2 = grads gc_is_none = gc is None gh_is_none = gh is None if gc_is_none: gc = 0 if gh_is_none: gh = 0 if ggc_prev1 is None: ggc_prev1 = 0 if ggc_prev2 is None: ggc_prev2 = 0 gc_prev1 = xp.empty_like(c_prev1) gc_prev2 = xp.empty_like(c_prev2) gx1 = xp.empty_like(x1) gx2 = xp.empty_like(x2) gc_next = xp.empty_like(c) ggc = xp.empty_like(c_prev1) ggh = xp.empty_like(c) a1, i1, f1, o1 = _extract_gates(x1) a2, i2, f2, o2 = _extract_gates(x2) gga1, ggi1, ggf1, ggo1 = _extract_gates(ggx1) gga2, ggi2, ggf2, ggo2 = _extract_gates(ggx2) ga1, gi1, gf1, go1 = _extract_gates(gx1) ga2, gi2, gf2, go2 = _extract_gates(gx2) o = o1 + o2 gc_prev1[:], ga1[:], gi1[:], gf1[:], go1[:], \ gc_prev2[:], ga2[:], gi2[:], gf2[:], go2[:], \ gc_next[:], ggc[:], ggh[:] \ = slstm_grad_grad(c_prev1, a1, i1, f1, c_prev2, a2, i2, f2, o, c, gc, gh, ggc_prev1, gga1, ggi1, ggf1, ggo1, ggc_prev2, gga2, ggi2, ggf2, ggo2) # If inputs were omitted, omit their gradients. if gc_is_none: ggc = None if gh_is_none: ggh = None return gc_prev1, gc_prev2, gx1, gx2, gc_next, ggc, ggh @cuda.fuse() def slstm_grad_grad(c_prev1, a1, i1, f1, c_prev2, a2, i2, f2, o, c, gc, gh, ggc_prev1, gga1, ggi1, ggf1, ggo1, ggc_prev2, gga2, ggi2, ggf2, ggo2): xp = backend.get_array_module(a1) sig_o = _sigmoid(o, xp) gsig_o = _grad_sigmoid(sig_o) ggsig_o = _grad_grad_sigmoid(sig_o) sig_i1 = _sigmoid(i1, xp) gsig_i1 = _grad_sigmoid(sig_i1) ggsig_i1 = _grad_grad_sigmoid(sig_i1) sig_i2 = _sigmoid(i2, xp) gsig_i2 = _grad_sigmoid(sig_i2) ggsig_i2 = _grad_grad_sigmoid(sig_i2) sig_f1 = _sigmoid(f1, xp) gsig_f1 = _grad_sigmoid(sig_f1) ggsig_f1 = _grad_grad_sigmoid(sig_f1) sig_f2 = _sigmoid(f2, xp) gsig_f2 = _grad_sigmoid(sig_f2) ggsig_f2 = _grad_grad_sigmoid(sig_f2) tanh_a1 = xp.tanh(a1) gtanh_a1 = _grad_tanh(tanh_a1) ggtanh_a1 = _grad_grad_tanh(tanh_a1, gtanh_a1) tanh_a2 = xp.tanh(a2) gtanh_a2 = _grad_tanh(tanh_a2) ggtanh_a2 = _grad_grad_tanh(tanh_a2, gtanh_a2) tanh_c = xp.tanh(c) gtanh_c = _grad_tanh(tanh_c) ggtanh_c = _grad_grad_tanh(tanh_c, gtanh_c) gc_bar = gh * sig_o * gtanh_c + gc gc_prev1 = ggf1 * gc_bar * gsig_f1 gc_prev2 = ggf2 * gc_bar * gsig_f2 ga1 = (gga1 * sig_i1 * ggtanh_a1 + ggi1 * gtanh_a1 * gsig_i1) * gc_bar ga2 = (gga2 * sig_i2 * ggtanh_a2 + ggi2 * gtanh_a2 * gsig_i2) * gc_bar gi1 = (gga1 * gtanh_a1 * gsig_i1 + ggi1 * tanh_a1 * ggsig_i1) * gc_bar gi2 = (gga2 * gtanh_a2 * gsig_i2 + ggi2 * tanh_a2 * ggsig_i2) * gc_bar gf1 = (ggc_prev1 * (gh * sig_o * gtanh_c + gc) * gsig_f1 + ggf1 * gc_bar * c_prev1 * ggsig_f1) gf2 = (ggc_prev2 * (gh * sig_o * gtanh_c + gc) * gsig_f2 + ggf2 * gc_bar * c_prev2 * ggsig_f2) ggc = ( ggc_prev1 * sig_f1 + gga1 * sig_i1 * gtanh_a1 + ggi1 * tanh_a1 * gsig_i1 + ggf1 * c_prev1 * gsig_f1 + ggc_prev2 * sig_f2 + gga2 * sig_i2 * gtanh_a2 + ggi2 * tanh_a2 * gsig_i2 + ggf2 * c_prev2 * gsig_f2) dgc_do = gh * gsig_o * gtanh_c go1 = go2 = ggc * dgc_do + (ggo1 + ggo2) * gh * tanh_c * ggsig_o dgc_dc = gh * sig_o * ggtanh_c gc_next = ggc * dgc_dc + (ggo1 + ggo2) * gh * gtanh_c * gsig_o ggh = ggc * sig_o * gtanh_c + (ggo1 + ggo2) * tanh_c * gsig_o return gc_prev1, ga1, gi1, gf1, go1, gc_prev2, ga2, gi2, gf2, go2, \ gc_next, ggc, ggh def slstm(c_prev1, c_prev2, x1, x2): """S-LSTM units as an activation function. This function implements S-LSTM unit. It is an extension of LSTM unit applied to tree structures. The function is applied to binary trees. Each node has two child nodes. It gets four arguments, previous cell states ``c_prev1`` and ``c_prev2``, and input arrays ``x1`` and ``x2``. First both input arrays ``x1`` and ``x2`` are split into eight arrays :math:`a_1, i_1, f_1, o_1`, and :math:`a_2, i_2, f_2, o_2`. They have the same shape along the second axis. It means that ``x1`` and ``x2`` 's second axis must have 4 times the length of ``c_prev1`` and ``c_prev2``. The split input arrays are corresponding to: - :math:`a_i` : sources of cell input - :math:`i_i` : sources of input gate - :math:`f_i` : sources of forget gate - :math:`o_i` : sources of output gate It computes the updated cell state ``c`` and the outgoing signal ``h`` as: .. math:: c &= \\tanh(a_1 + a_2) \\sigma(i_1 + i_2) + c_{\\text{prev}1} \\sigma(f_1) + c_{\\text{prev}2} \\sigma(f_2), \\\\ h &= \\tanh(c) \\sigma(o_1 + o_2), where :math:`\\sigma` is the elementwise sigmoid function. The function returns ``c`` and ``h`` as a tuple. Args: c_prev1 (:class:`~chainer.Variable` or :ref:`ndarray`): Variable that holds the previous cell state of the first child node. The cell state should be a zero array or the output of the previous call of LSTM. c_prev2 (:class:`~chainer.Variable` or :ref:`ndarray`): Variable that holds the previous cell state of the second child node. x1 (:class:`~chainer.Variable` or :ref:`ndarray`): Variable that holds the sources of cell input, input gate, forget gate and output gate from the first child node. It must have the second dimension whose size is four times of that of the cell state. x2 (:class:`~chainer.Variable` or :ref:`ndarray`): Variable that holds the input sources from the second child node. Returns: tuple: Two :class:`~chainer.Variable` objects ``c`` and ``h``. ``c`` is the cell state. ``h`` indicates the outgoing signal. See detail in paper: `Long Short-Term Memory Over Tree Structures <https://arxiv.org/abs/1503.04881>`_. .. admonition:: Example Assuming ``c1``, ``c2`` is the previous cell state of children, and ``h1``, ``h2`` is the previous outgoing signal from children. Each of ``c1``, ``c2``, ``h1`` and ``h2`` has ``n_units`` channels. Most typical preparation of ``x1``, ``x2`` is: >>> n_units = 100 >>> h1 = chainer.Variable(np.zeros((1, n_units), np.float32)) >>> h2 = chainer.Variable(np.zeros((1, n_units), np.float32)) >>> c1 = chainer.Variable(np.zeros((1, n_units), np.float32)) >>> c2 = chainer.Variable(np.zeros((1, n_units), np.float32)) >>> model1 = chainer.Chain() >>> with model1.init_scope(): ... model1.w = L.Linear(n_units, 4 * n_units) ... model1.v = L.Linear(n_units, 4 * n_units) >>> model2 = chainer.Chain() >>> with model2.init_scope(): ... model2.w = L.Linear(n_units, 4 * n_units) ... model2.v = L.Linear(n_units, 4 * n_units) >>> x1 = model1.w(c1) + model1.v(h1) >>> x2 = model2.w(c2) + model2.v(h2) >>> c, h = F.slstm(c1, c2, x1, x2) It corresponds to calculate the input array ``x1``, or the input sources :math:`a_1, i_1, f_1, o_1` from the previous cell state of first child node ``c1``, and the previous outgoing signal from first child node ``h1``. Different parameters are used for different kind of input sources. """ return SLSTM().apply((c_prev1, c_prev2, x1, x2))
15,311
34.526682
79
py
chainer
chainer-master/chainer/functions/rnn/n_step_lstm.py
import numpy import chainer from chainer import backend from chainer.backends import cuda from chainer.functions.array import reshape from chainer.functions.array import stack from chainer.functions.connection import linear from chainer.functions.rnn import lstm from chainer.functions.rnn import n_step_rnn from chainer.utils import argument from chainer import variable import chainerx if cuda.cudnn_enabled: cudnn = cuda.cudnn def _extract_apply_in_data(inputs): if not inputs: return False, () if chainerx.is_available(): has_chainerx_array = False # Unwrap arrays arrays = [] for x in inputs: if isinstance(x, variable.Variable): if x._has_chainerx_array: arrays.append(x._data[0]) has_chainerx_array = True else: arrays.append(x.array) else: # x is ndarray arrays.append(x) if not has_chainerx_array: if isinstance(x, chainerx.ndarray): has_chainerx_array = True return has_chainerx_array, tuple(arrays) else: return False, tuple([ x.array if isinstance(x, variable.Variable) else x for x in inputs]) def _combine_inputs(hx, cx, ws, bs, xs, num_layers, directions): combined = [] combined.append(hx) combined.append(cx) for x in xs: combined.append(x) for n in range(num_layers): for direction in range(directions): idx = directions * n + direction for i in range(8): combined.append(ws[idx][i]) for i in range(8): combined.append(bs[idx][i]) return combined def _seperate_inputs(combined, num_layers, seq_length, directions): hx = combined[0] cx = combined[1] xs = combined[2: 2 + seq_length] ws = [] bs = [] index = 2 + seq_length for n in range(num_layers): ws.append(combined[index: index + 8]) bs.append(combined[index + 8: index + 16]) index += 16 if directions == 2: ws.append(combined[index: index + 8]) bs.append(combined[index + 8: index + 16]) index += 16 return hx, cx, ws, bs, xs def _stack_weight(ws): # TODO(unno): Input of the current LSTM implementation is shuffled w = stack.stack(ws, axis=1) shape = w.shape return reshape.reshape(w, (shape[0] * shape[1],) + shape[2:]) class NStepLSTM(n_step_rnn.BaseNStepRNN): def __init__(self, n_layers, states, lengths): n_step_rnn.BaseNStepRNN.__init__( self, n_layers, states, lengths, rnn_dir='uni', rnn_mode='lstm') class NStepBiLSTM(n_step_rnn.BaseNStepRNN): def __init__(self, n_layers, states, lengths): n_step_rnn.BaseNStepRNN.__init__( self, n_layers, states, lengths, rnn_dir='bi', rnn_mode='lstm') def n_step_lstm( n_layers, dropout_ratio, hx, cx, ws, bs, xs, **kwargs): """n_step_lstm(n_layers, dropout_ratio, hx, cx, ws, bs, xs) Stacked Uni-directional Long Short-Term Memory function. This function calculates stacked Uni-directional LSTM with sequences. This function gets an initial hidden state :math:`h_0`, an initial cell state :math:`c_0`, an input sequence :math:`x`, weight matrices :math:`W`, and bias vectors :math:`b`. This function calculates hidden states :math:`h_t` and :math:`c_t` for each time :math:`t` from input :math:`x_t`. .. math:: i_t &= \\sigma(W_0 x_t + W_4 h_{t-1} + b_0 + b_4) \\\\ f_t &= \\sigma(W_1 x_t + W_5 h_{t-1} + b_1 + b_5) \\\\ o_t &= \\sigma(W_2 x_t + W_6 h_{t-1} + b_2 + b_6) \\\\ a_t &= \\tanh(W_3 x_t + W_7 h_{t-1} + b_3 + b_7) \\\\ c_t &= f_t \\cdot c_{t-1} + i_t \\cdot a_t \\\\ h_t &= o_t \\cdot \\tanh(c_t) As the function accepts a sequence, it calculates :math:`h_t` for all :math:`t` with one call. Eight weight matrices and eight bias vectors are required for each layer. So, when :math:`S` layers exist, you need to prepare :math:`8S` weight matrices and :math:`8S` bias vectors. If the number of layers ``n_layers`` is greater than :math:`1`, the input of the ``k``-th layer is the hidden state ``h_t`` of the ``k-1``-th layer. Note that all input variables except the first layer may have different shape from the first layer. Args: n_layers(int): The number of layers. dropout_ratio(float): Dropout ratio. hx (:class:`~chainer.Variable`): Variable holding stacked hidden states. Its shape is ``(S, B, N)`` where ``S`` is the number of layers and is equal to ``n_layers``, ``B`` is the mini-batch size, and ``N`` is the dimension of the hidden units. cx (:class:`~chainer.Variable`): Variable holding stacked cell states. It has the same shape as ``hx``. ws (list of list of :class:`~chainer.Variable`): Weight matrices. ``ws[i]`` represents the weights for the i-th layer. Each ``ws[i]`` is a list containing eight matrices. ``ws[i][j]`` corresponds to :math:`W_j` in the equation. Only ``ws[0][j]`` where ``0 <= j < 4`` are ``(N, I)``-shaped as they are multiplied with input variables, where ``I`` is the size of the input and ``N`` is the dimension of the hidden units. All other matrices are ``(N, N)``-shaped. bs (list of list of :class:`~chainer.Variable`): Bias vectors. ``bs[i]`` represents the biases for the i-th layer. Each ``bs[i]`` is a list containing eight vectors. ``bs[i][j]`` corresponds to :math:`b_j` in the equation. The shape of each matrix is ``(N,)`` where ``N`` is the dimension of the hidden units. xs (list of :class:`~chainer.Variable`): A list of :class:`~chainer.Variable` holding input values. Each element ``xs[t]`` holds input value for time ``t``. Its shape is ``(B_t, I)``, where ``B_t`` is the mini-batch size for time ``t``. The sequences must be transposed. :func:`~chainer.functions.transpose_sequence` can be used to transpose a list of :class:`~chainer.Variable`\\ s each representing a sequence. When sequences has different lengths, they must be sorted in descending order of their lengths before transposing. So ``xs`` needs to satisfy ``xs[t].shape[0] >= xs[t + 1].shape[0]``. Returns: tuple: This function returns a tuple containing three elements, ``hy``, ``cy`` and ``ys``. - ``hy`` is an updated hidden states whose shape is the same as ``hx``. - ``cy`` is an updated cell states whose shape is the same as ``cx``. - ``ys`` is a list of :class:`~chainer.Variable` . Each element ``ys[t]`` holds hidden states of the last layer corresponding to an input ``xs[t]``. Its shape is ``(B_t, N)`` where ``B_t`` is the mini-batch size for time ``t``, and ``N`` is size of hidden units. Note that ``B_t`` is the same value as ``xs[t]``. .. note:: The dimension of hidden units is limited to only one size ``N``. If you want to use variable dimension of hidden units, please use :class:`chainer.functions.lstm`. .. seealso:: :func:`chainer.functions.lstm` .. admonition:: Example >>> batchs = [3, 2, 1] # support variable length sequences >>> in_size, out_size, n_layers = 3, 2, 2 >>> dropout_ratio = 0.0 >>> xs = [np.ones((b, in_size)).astype(np.float32) for b in batchs] >>> [x.shape for x in xs] [(3, 3), (2, 3), (1, 3)] >>> h_shape = (n_layers, batchs[0], out_size) >>> hx = np.ones(h_shape).astype(np.float32) >>> cx = np.ones(h_shape).astype(np.float32) >>> w_in = lambda i, j: in_size if i == 0 and j < 4 else out_size >>> ws = [] >>> bs = [] >>> for n in range(n_layers): ... ws.append([np.ones((out_size, w_in(n, i))).astype(np.float32) \ for i in range(8)]) ... bs.append([np.ones((out_size,)).astype(np.float32) \ for _ in range(8)]) ... >>> ws[0][0].shape # ws[0][:4].shape are (out_size, in_size) (2, 3) >>> ws[1][0].shape # others are (out_size, out_size) (2, 2) >>> bs[0][0].shape (2,) >>> hy, cy, ys = F.n_step_lstm( ... n_layers, dropout_ratio, hx, cx, ws, bs, xs) >>> hy.shape (2, 3, 2) >>> cy.shape (2, 3, 2) >>> [y.shape for y in ys] [(3, 2), (2, 2), (1, 2)] """ return n_step_lstm_base(n_layers, dropout_ratio, hx, cx, ws, bs, xs, use_bi_direction=False, **kwargs) def n_step_bilstm( n_layers, dropout_ratio, hx, cx, ws, bs, xs, **kwargs): """n_step_bilstm(n_layers, dropout_ratio, hx, cx, ws, bs, xs) Stacked Bi-directional Long Short-Term Memory function. This function calculates stacked Bi-directional LSTM with sequences. This function gets an initial hidden state :math:`h_0`, an initial cell state :math:`c_0`, an input sequence :math:`x`, weight matrices :math:`W`, and bias vectors :math:`b`. This function calculates hidden states :math:`h_t` and :math:`c_t` for each time :math:`t` from input :math:`x_t`. .. math:: i^{f}_t &=& \\sigma(W^{f}_0 x_t + W^{f}_4 h_{t-1} + b^{f}_0 + b^{f}_4), \\\\ f^{f}_t &=& \\sigma(W^{f}_1 x_t + W^{f}_5 h_{t-1} + b^{f}_1 + b^{f}_5), \\\\ o^{f}_t &=& \\sigma(W^{f}_2 x_t + W^{f}_6 h_{t-1} + b^{f}_2 + b^{f}_6), \\\\ a^{f}_t &=& \\tanh(W^{f}_3 x_t + W^{f}_7 h_{t-1} + b^{f}_3 + b^{f}_7), \\\\ c^{f}_t &=& f^{f}_t \\cdot c^{f}_{t-1} + i^{f}_t \\cdot a^{f}_t, \\\\ h^{f}_t &=& o^{f}_t \\cdot \\tanh(c^{f}_t), \\\\ i^{b}_t &=& \\sigma(W^{b}_0 x_t + W^{b}_4 h_{t-1} + b^{b}_0 + b^{b}_4), \\\\ f^{b}_t &=& \\sigma(W^{b}_1 x_t + W^{b}_5 h_{t-1} + b^{b}_1 + b^{b}_5), \\\\ o^{b}_t &=& \\sigma(W^{b}_2 x_t + W^{b}_6 h_{t-1} + b^{b}_2 + b^{b}_6), \\\\ a^{b}_t &=& \\tanh(W^{b}_3 x_t + W^{b}_7 h_{t-1} + b^{b}_3 + b^{b}_7), \\\\ c^{b}_t &=& f^{b}_t \\cdot c^{b}_{t-1} + i^{b}_t \\cdot a^{b}_t, \\\\ h^{b}_t &=& o^{b}_t \\cdot \\tanh(c^{b}_t), \\\\ h_t &=& [h^{f}_t; h^{b}_t] where :math:`W^{f}` is the weight matrices for forward-LSTM, :math:`W^{b}` is weight matrices for backward-LSTM. As the function accepts a sequence, it calculates :math:`h_t` for all :math:`t` with one call. Eight weight matrices and eight bias vectors are required for each layer of each direction. So, when :math:`S` layers exist, you need to prepare :math:`16S` weight matrices and :math:`16S` bias vectors. If the number of layers ``n_layers`` is greater than :math:`1`, the input of the ``k``-th layer is the hidden state ``h_t`` of the ``k-1``-th layer. Note that all input variables except the first layer may have different shape from the first layer. Args: n_layers(int): The number of layers. dropout_ratio(float): Dropout ratio. hx (:class:`~chainer.Variable`): Variable holding stacked hidden states. Its shape is ``(2S, B, N)`` where ``S`` is the number of layers and is equal to ``n_layers``, ``B`` is the mini-batch size, and ``N`` is the dimension of the hidden units. Because of bi-direction, the first dimension length is ``2S``. cx (:class:`~chainer.Variable`): Variable holding stacked cell states. It has the same shape as ``hx``. ws (list of list of :class:`~chainer.Variable`): Weight matrices. ``ws[2 * l + m]`` represents the weights for the l-th layer of the m-th direction. (``m == 0`` means the forward direction and ``m == 1`` means the backward direction.) Each ``ws[i]`` is a list containing eight matrices. ``ws[i][j]`` corresponds to :math:`W_j` in the equation. ``ws[0][j]`` and ``ws[1][j]`` where ``0 <= j < 4`` are ``(N, I)``-shaped because they are multiplied with input variables, where ``I`` is the size of the input. ``ws[i][j]`` where ``2 <= i`` and ``0 <= j < 4`` are ``(N, 2N)``-shaped because they are multiplied with two hidden layers :math:`h_t = [h^{f}_t; h^{b}_t]`. All other matrices are ``(N, N)``-shaped. bs (list of list of :class:`~chainer.Variable`): Bias vectors. ``bs[2 * l + m]`` represents the weights for the l-th layer of m-th direction. (``m == 0`` means the forward direction and ``m == 1`` means the backward direction.) Each ``bs[i]`` is a list containing eight vectors. ``bs[i][j]`` corresponds to :math:`b_j` in the equation. The shape of each matrix is ``(N,)``. xs (list of :class:`~chainer.Variable`): A list of :class:`~chainer.Variable` holding input values. Each element ``xs[t]`` holds input value for time ``t``. Its shape is ``(B_t, I)``, where ``B_t`` is the mini-batch size for time ``t``. The sequences must be transposed. :func:`~chainer.functions.transpose_sequence` can be used to transpose a list of :class:`~chainer.Variable`\\ s each representing a sequence. When sequences has different lengths, they must be sorted in descending order of their lengths before transposing. So ``xs`` needs to satisfy ``xs[t].shape[0] >= xs[t + 1].shape[0]``. Returns: tuple: This function returns a tuple containing three elements, ``hy``, ``cy`` and ``ys``. - ``hy`` is an updated hidden states whose shape is the same as ``hx``. - ``cy`` is an updated cell states whose shape is the same as ``cx``. - ``ys`` is a list of :class:`~chainer.Variable` . Each element ``ys[t]`` holds hidden states of the last layer corresponding to an input ``xs[t]``. Its shape is ``(B_t, 2N)`` where ``B_t`` is the mini-batch size for time ``t``, and ``N`` is size of hidden units. Note that ``B_t`` is the same value as ``xs[t]``. .. admonition:: Example >>> batchs = [3, 2, 1] # support variable length sequences >>> in_size, out_size, n_layers = 3, 2, 2 >>> dropout_ratio = 0.0 >>> xs = [np.ones((b, in_size)).astype(np.float32) for b in batchs] >>> [x.shape for x in xs] [(3, 3), (2, 3), (1, 3)] >>> h_shape = (n_layers * 2, batchs[0], out_size) >>> hx = np.ones(h_shape).astype(np.float32) >>> cx = np.ones(h_shape).astype(np.float32) >>> def w_in(i, j): ... if i == 0 and j < 4: ... return in_size ... elif i > 0 and j < 4: ... return out_size * 2 ... else: ... return out_size ... >>> ws = [] >>> bs = [] >>> for n in range(n_layers): ... for direction in (0, 1): ... ws.append([np.ones((out_size, w_in(n, i))).\ astype(np.float32) for i in range(8)]) ... bs.append([np.ones((out_size,)).astype(np.float32) \ for _ in range(8)]) ... >>> ws[0][0].shape # ws[0:2][:4].shape are (out_size, in_size) (2, 3) >>> ws[2][0].shape # ws[2:][:4].shape are (out_size, 2 * out_size) (2, 4) >>> ws[0][4].shape # others are (out_size, out_size) (2, 2) >>> bs[0][0].shape (2,) >>> hy, cy, ys = F.n_step_bilstm( ... n_layers, dropout_ratio, hx, cx, ws, bs, xs) >>> hy.shape (4, 3, 2) >>> cy.shape (4, 3, 2) >>> [y.shape for y in ys] [(3, 4), (2, 4), (1, 4)] """ return n_step_lstm_base(n_layers, dropout_ratio, hx, cx, ws, bs, xs, use_bi_direction=True, **kwargs) def n_step_lstm_base( n_layers, dropout_ratio, hx, cx, ws, bs, xs, use_bi_direction, **kwargs): """Base function for Stack LSTM/BiLSTM functions. This function is used at :func:`chainer.functions.n_step_lstm` and :func:`chainer.functions.n_step_bilstm`. This function's behavior depends on following arguments, ``activation`` and ``use_bi_direction``. Args: n_layers(int): The number of layers. dropout_ratio(float): Dropout ratio. hx (:class:`~chainer.Variable`): Variable holding stacked hidden states. Its shape is ``(S, B, N)`` where ``S`` is the number of layers and is equal to ``n_layers``, ``B`` is the mini-batch size, and ``N`` is the dimension of the hidden units. cx (:class:`~chainer.Variable`): Variable holding stacked cell states. It has the same shape as ``hx``. ws (list of list of :class:`~chainer.Variable`): Weight matrices. ``ws[i]`` represents the weights for the i-th layer. Each ``ws[i]`` is a list containing eight matrices. ``ws[i][j]`` corresponds to :math:`W_j` in the equation. Only ``ws[0][j]`` where ``0 <= j < 4`` are ``(N, I)``-shape as they are multiplied with input variables, where ``I`` is the size of the input and ``N`` is the dimension of the hidden units. All other matrices are ``(N, N)``-shaped. bs (list of list of :class:`~chainer.Variable`): Bias vectors. ``bs[i]`` represents the biases for the i-th layer. Each ``bs[i]`` is a list containing eight vectors. ``bs[i][j]`` corresponds to :math:`b_j` in the equation. The shape of each matrix is ``(N,)``. xs (list of :class:`~chainer.Variable`): A list of :class:`~chainer.Variable` holding input values. Each element ``xs[t]`` holds input value for time ``t``. Its shape is ``(B_t, I)``, where ``B_t`` is the mini-batch size for time ``t``. The sequences must be transposed. :func:`~chainer.functions.transpose_sequence` can be used to transpose a list of :class:`~chainer.Variable`\\ s each representing a sequence. When sequences has different lengths, they must be sorted in descending order of their lengths before transposing. So ``xs`` needs to satisfy ``xs[t].shape[0] >= xs[t + 1].shape[0]``. use_bi_direction (bool): If ``True``, this function uses Bi-directional LSTM. Returns: tuple: This function returns a tuple containing three elements, ``hy``, ``cy`` and ``ys``. - ``hy`` is an updated hidden states whose shape is the same as ``hx``. - ``cy`` is an updated cell states whose shape is the same as ``cx``. - ``ys`` is a list of :class:`~chainer.Variable` . Each element ``ys[t]`` holds hidden states of the last layer corresponding to an input ``xs[t]``. Its shape is ``(B_t, N)`` where ``B_t`` is the mini-batch size for time ``t``. Note that ``B_t`` is the same value as ``xs[t]``. .. seealso:: :func:`chainer.functions.n_step_lstm` :func:`chainer.functions.n_step_bilstm` """ if kwargs: argument.check_unexpected_kwargs( kwargs, train='train argument is not supported anymore. ' 'Use chainer.using_config', use_cudnn='use_cudnn argument is not supported anymore. ' 'Use chainer.using_config') argument.assert_kwargs_empty(kwargs) # Check input size consistency with xs and ws here. x_in = xs[0].shape[1] w_in = ws[0][0].shape[1] if x_in != w_in: raise ValueError('Inconsistent input size in input values and weight ' 'parameters: {} != {}'.format(x_in, w_in)) xp = backend.get_array_module(hx, hx.data) use_cuda = xp is cuda.cupy or ( xp is chainerx and hx.device.device.backend.name == 'cuda') directions = 1 if use_bi_direction: directions = 2 combined = _combine_inputs(hx, cx, ws, bs, xs, n_layers, directions) has_chainerx_array, combined = _extract_apply_in_data(combined) hx_chx, cx_chx, ws_chx, bs_chx, xs_chx = _seperate_inputs( combined, n_layers, len(xs), directions) if has_chainerx_array and xp is chainerx and dropout_ratio == 0: if use_bi_direction: hy, cy, ys = chainerx.n_step_bilstm( n_layers, hx_chx, cx_chx, ws_chx, bs_chx, xs_chx) else: hy, cy, ys = chainerx.n_step_lstm( n_layers, hx_chx, cx_chx, ws_chx, bs_chx, xs_chx) hy = variable.Variable._init_unchecked( hy, requires_grad=hy.is_backprop_required(), is_chainerx_array=True) cy = variable.Variable._init_unchecked( cy, requires_grad=cy.is_backprop_required(), is_chainerx_array=True) ys = [variable.Variable._init_unchecked( y, requires_grad=y.is_backprop_required(), is_chainerx_array=True) for y in ys] return hy, cy, ys elif use_cuda and chainer.should_use_cudnn('>=auto', 5000): lengths = [len(x) for x in xs] xs = chainer.functions.concat(xs, axis=0) with chainer.using_device(xs.device): states = cuda.get_cudnn_dropout_states() states.set_dropout_ratio(dropout_ratio) w = n_step_rnn.cudnn_rnn_weight_concat( n_layers, states, use_bi_direction, 'lstm', ws, bs) if use_bi_direction: rnn = NStepBiLSTM else: rnn = NStepLSTM hy, cy, ys = rnn(n_layers, states, lengths)(hx, cx, w, xs) sections = numpy.cumsum(lengths[:-1]) ys = chainer.functions.split_axis(ys, sections, 0) return hy, cy, ys else: return n_step_rnn.n_step_rnn_impl( _lstm, n_layers, dropout_ratio, hx, cx, ws, bs, xs, use_bi_direction) def _lstm(x, h, c, w, b): xw = _stack_weight([w[2], w[0], w[1], w[3]]) hw = _stack_weight([w[6], w[4], w[5], w[7]]) xb = _stack_weight([b[2], b[0], b[1], b[3]]) hb = _stack_weight([b[6], b[4], b[5], b[7]]) lstm_in = linear.linear(x, xw, xb) + linear.linear(h, hw, hb) c_bar, h_bar = lstm.lstm(c, lstm_in) return h_bar, c_bar
22,970
40.917883
79
py
chainer
chainer-master/chainer/functions/rnn/__init__.py
0
0
0
py
chainer
chainer-master/chainer/functions/rnn/n_step_rnn.py
import itertools import numpy import six import chainer import chainerx from chainer import backend from chainer import variable from chainer.backends import cuda from chainer import configuration from chainer import function from chainer.functions.activation import relu from chainer.functions.activation import tanh from chainer.functions.array import concat from chainer.functions.array import split_axis from chainer.functions.array import stack from chainer.functions.connection import linear from chainer.functions.noise import dropout from chainer.utils import argument from chainer.utils import type_check if cuda.cudnn_enabled: cudnn = cuda.cudnn libcudnn = cuda.libcudnn _cudnn_version = libcudnn.getVersion() if cuda.cudnn_enabled and _cudnn_version >= 5000: # Define RNN parameters using dict. _rnn_dirs = { 'uni': libcudnn.CUDNN_UNIDIRECTIONAL, 'bi': libcudnn.CUDNN_BIDIRECTIONAL, } _rnn_modes = { 'rnn_relu': libcudnn.CUDNN_RNN_RELU, 'rnn_tanh': libcudnn.CUDNN_RNN_TANH, 'gru': libcudnn.CUDNN_GRU, 'lstm': libcudnn.CUDNN_LSTM, } _rnn_n_params = { libcudnn.CUDNN_RNN_RELU: 2, libcudnn.CUDNN_RNN_TANH: 2, libcudnn.CUDNN_GRU: 6, libcudnn.CUDNN_LSTM: 8, } _rnn_params_direction = { libcudnn.CUDNN_UNIDIRECTIONAL: 1, libcudnn.CUDNN_BIDIRECTIONAL: 2, } _rnn_params_use_cell = { libcudnn.CUDNN_RNN_RELU: False, libcudnn.CUDNN_RNN_TANH: False, libcudnn.CUDNN_GRU: False, libcudnn.CUDNN_LSTM: True, } def _extract_apply_in_data(inputs): if not inputs: return False, () if chainerx.is_available(): has_chainerx_array = False # Unwrap arrays arrays = [] for x in inputs: if isinstance(x, variable.Variable): if x._has_chainerx_array: arrays.append(x._data[0]) has_chainerx_array = True else: arrays.append(x.array) else: # x is ndarray arrays.append(x) if not has_chainerx_array: if isinstance(x, chainerx.ndarray): has_chainerx_array = True return has_chainerx_array, tuple(arrays) else: return False, tuple([ x.array if isinstance(x, variable.Variable) else x for x in inputs]) def _combine_inputs(hx, ws, bs, xs, num_layers, directions): combined = [] combined.append(hx) for x in xs: combined.append(x) for n in range(num_layers): for direction in range(directions): idx = directions * n + direction for i in range(2): combined.append(ws[idx][i]) for i in range(2): combined.append(bs[idx][i]) return combined def _seperate_inputs(combined, num_layers, seq_length, directions): hx = combined[0] xs = combined[1: 1 + seq_length] ws = [] bs = [] index = 1 + seq_length for n in range(num_layers): ws.append(combined[index: index + 2]) bs.append(combined[index + 2: index + 4]) index += 4 if directions == 2: ws.append(combined[index: index + 2]) bs.append(combined[index + 2: index + 4]) index += 4 return hx, ws, bs, xs class CudnnRNNWeightConcat(function.Function): """Concatenates weight matrices for cuDNN's RNN. This function concatenates weight matrices for RNNs into one large array. Its format is defined in cuDNN's API. """ def __init__(self, n_layers, states, rnn_dir, rnn_mode): self.n_layers = n_layers self.states = states self.rnn_dir = _rnn_dirs[rnn_dir] self.rnn_mode = _rnn_modes[rnn_mode] self.rnn_direction = _rnn_params_direction[self.rnn_dir] self.n_W = _rnn_n_params[self.rnn_mode] def check_type_forward(self, in_types): n_params = self.n_layers * self.rnn_direction * self.n_W type_check.expect( in_types.size() == n_params * 2) w_types = in_types[:n_params] b_types = in_types[n_params:] in_size = w_types[0].shape[1] out_size = w_types[0].shape[0] dtype = w_types[0].dtype type_check.expect(dtype.kind == 'f') for layer in six.moves.range(self.n_layers): for di in six.moves.range(self.rnn_direction): for i in six.moves.range(self.n_W): ind = (layer * self.rnn_direction + di) * self.n_W + i w_type = w_types[ind] b_type = b_types[ind] if self.rnn_direction == 1: # Uni-direction if layer == 0 and i < (self.n_W // 2): w_in = in_size else: w_in = out_size else: # Bi-direction if layer == 0 and i < (self.n_W // 2): w_in = in_size elif layer > 0 and i < (self.n_W // 2): w_in = out_size * self.rnn_direction else: w_in = out_size type_check.expect( w_type.dtype == dtype, w_type.ndim == 2, w_type.shape[0] == out_size, w_type.shape[1] == w_in, b_type.dtype == dtype, b_type.ndim == 1, b_type.shape[0] == out_size, ) def forward_gpu(self, inputs): handle = cudnn.get_handle() ws_size = self.n_layers * self.rnn_direction * self.n_W ws = inputs[:ws_size] bs = inputs[ws_size:] out_size = ws[0].shape[0] in_size = ws[0].shape[1] dtype = ws[0].dtype cudnn_data_type = cudnn.get_data_type(dtype) # TODO(unno): Make a wrapper method to avoid access _desc directly rnn_desc = cudnn.create_rnn_descriptor( out_size, self.n_layers, self.states._desc, libcudnn.CUDNN_LINEAR_INPUT, self.rnn_dir, self.rnn_mode, cudnn_data_type) self.rnn_desc = rnn_desc dummy_x = cuda.cupy.empty((1, in_size, 1), dtype=dtype) x_desc = cudnn.create_tensor_nd_descriptor(dummy_x) weights_size = libcudnn.getRNNParamsSize( handle, rnn_desc.value, x_desc.value, cudnn_data_type) byte_size = dtype.itemsize w = cuda.cupy.empty((weights_size // byte_size, 1, 1), dtype=dtype) w_desc = cudnn.create_filter_descriptor(w) for layer in six.moves.range(self.n_layers): for di in six.moves.range(self.rnn_direction): mat_index = layer * self.rnn_direction + di # di = 0: forward, 1: backward for lin_layer_id in six.moves.range(self.n_W): mat = cudnn.get_rnn_lin_layer_matrix_params( handle, rnn_desc, mat_index, x_desc, w_desc, w, lin_layer_id) W_index = mat_index * self.n_W + lin_layer_id m = mat.reshape(mat.size) m[...] = ws[W_index].ravel() bias = cudnn.get_rnn_lin_layer_bias_params( handle, rnn_desc, mat_index, x_desc, w_desc, w, lin_layer_id) b = bias.reshape(bias.size) b[...] = bs[W_index] self.w_desc = w_desc self.x_desc = x_desc return w, def backward(self, inputs, grads): handle = cudnn.get_handle() ws_size = self.n_layers * self.rnn_direction * self.n_W ws = inputs[0:ws_size] bs = inputs[ws_size:] rnn_desc = self.rnn_desc dw = grads[0] dw_desc = cudnn.create_filter_descriptor(dw) dx_desc = self.x_desc dws = [] dbs = [] for layer in six.moves.range(self.n_layers): for di in six.moves.range(self.rnn_direction): mat_index = layer * self.rnn_direction + di for lin_layer_id in six.moves.range(self.n_W): mat = cudnn.get_rnn_lin_layer_matrix_params( handle, rnn_desc, mat_index, dx_desc, dw_desc, dw, lin_layer_id) W_index = mat_index * self.n_W + lin_layer_id dws.append(mat.reshape(ws[W_index].shape)) bias = cudnn.get_rnn_lin_layer_bias_params( handle, rnn_desc, mat_index, dx_desc, dw_desc, dw, lin_layer_id) dbs.append(bias.reshape(bs[W_index].shape)) return tuple(dws + dbs) def cudnn_rnn_weight_concat( n_layers, states, use_bi_direction, rnn_mode, ws, bs): rnn_dir = 'bi' if use_bi_direction else 'uni' inputs = itertools.chain( itertools.chain.from_iterable(ws), itertools.chain.from_iterable(bs), ) return CudnnRNNWeightConcat(n_layers, states, rnn_dir, rnn_mode)(*inputs) class BaseNStepRNN(function.Function): def __init__(self, n_layers, states, lengths, rnn_dir, rnn_mode, **kwargs): if kwargs: argument.check_unexpected_kwargs( kwargs, train='train argument is not supported anymore. ' 'Use chainer.using_config') argument.assert_kwargs_empty(kwargs) if rnn_dir not in _rnn_dirs: candidate_list = ','.join(_rnn_dirs.keys()) raise ValueError('Invalid rnn_dir: "%s". Please select from [%s]' % (rnn_dir, candidate_list)) if rnn_mode not in _rnn_modes: candidate_list = ','.join(_rnn_modes.keys()) raise ValueError('Invalid rnn_mode: "%s". Please select from [%s]' % (rnn_mode, candidate_list)) self.rnn_dir = _rnn_dirs[rnn_dir] self.rnn_mode = _rnn_modes[rnn_mode] self.rnn_direction = _rnn_params_direction[self.rnn_dir] self.n_layers = n_layers self.states = states self.use_cell = _rnn_params_use_cell[self.rnn_mode] self.lengths = lengths self.sections = numpy.cumsum(lengths) def check_type_forward(self, in_types): if self.use_cell: type_check.expect(in_types.size() == 4) h_type, c_type, w_type, x_type = in_types h_size = self.n_layers * self.rnn_direction type_check.expect( h_type.dtype == x_type.dtype, c_type.dtype == x_type.dtype, h_type.ndim == 3, h_type.shape[0] == h_size, c_type.ndim == 3, c_type.shape[0] == h_size, # mini-batch size h_type.shape[1] == c_type.shape[1], # hidden size h_type.shape[2] == c_type.shape[2], ) else: type_check.expect(in_types.size() == 3) h_type, w_type, x_type = in_types h_size = self.n_layers * self.rnn_direction type_check.expect( h_type.dtype == x_type.dtype, h_type.ndim == 3, h_type.shape[0] == h_size, ) type_check.expect( x_type.dtype.kind == 'f', x_type.ndim == 2, x_type.shape[0] == self.sections[-1], ) def forward_gpu(self, inputs): if self.use_cell: # LSTM hx, cx, w, xs = inputs else: # RNN, GRU hx, w, xs = inputs cx = None if not configuration.config.train: hy, cy, ys = cudnn.rnn_forward_inference( self.states, self.rnn_dir, self.rnn_mode, hx, cx, w, xs, self.lengths) else: self.reserve_space, hy, cy, ys = cudnn.rnn_forward_training( self.states, self.rnn_dir, self.rnn_mode, hx, cx, w, xs, self.lengths) if self.use_cell: # LSTM self.retain_outputs((2,)) return hy, cy, ys else: # GRU, RNN self.retain_outputs((1,)) return hy, ys def backward(self, inputs, grads): if not configuration.config.train: raise RuntimeError('cuDNN does not support backward computation ' 'of RNN in testing mode') if self.use_cell: # LSTM hx, cx, w, xs = inputs dhy, dcy, dys = grads if dcy is None: dcy = cuda.cupy.zeros_like(cx) else: # GRU, RNN hx, w, xs = inputs dhy, dys = grads dcy = cx = None ys = self.output_data[-1] if dhy is None: dhy = cuda.cupy.zeros_like(hx) if dys is None: dys = cuda.cupy.zeros_like(ys) dhx, dcx, dxs = cudnn.rnn_backward_data( self.states, self.rnn_dir, self.rnn_mode, hx, cx, w, xs, ys, self.reserve_space, dhy, dcy, dys, self.lengths) dw = cudnn.rnn_backward_weights( self.states, self.rnn_dir, self.rnn_mode, xs, hx, ys, w, self.reserve_space, self.lengths) if self.use_cell: # LSTM return dhx, dcx, dw, dxs else: # GRU, RNN return dhx, dw, dxs class NStepRNNTanh(BaseNStepRNN): def __init__(self, n_layers, states, lengths, **kwargs): BaseNStepRNN.__init__( self, n_layers, states, lengths, rnn_dir='uni', rnn_mode='rnn_tanh', **kwargs) class NStepRNNReLU(BaseNStepRNN): def __init__(self, n_layers, states, lengths, **kwargs): BaseNStepRNN.__init__( self, n_layers, states, lengths, rnn_dir='uni', rnn_mode='rnn_relu', **kwargs) class NStepBiRNNTanh(BaseNStepRNN): def __init__(self, n_layers, states, lengths, **kwargs): BaseNStepRNN.__init__( self, n_layers, states, lengths, rnn_dir='bi', rnn_mode='rnn_tanh', **kwargs) class NStepBiRNNReLU(BaseNStepRNN): def __init__(self, n_layers, states, lengths, **kwargs): BaseNStepRNN.__init__( self, n_layers, states, lengths, rnn_dir='bi', rnn_mode='rnn_relu', **kwargs) def n_step_rnn( n_layers, dropout_ratio, hx, ws, bs, xs, activation='tanh', **kwargs): """n_step_rnn(n_layers, dropout_ratio, hx, ws, bs, xs, activation='tanh') Stacked Uni-directional RNN function for sequence inputs. This function calculates stacked Uni-directional RNN with sequences. This function gets an initial hidden state :math:`h_0`, an initial cell state :math:`c_0`, an input sequence :math:`x`, weight matrices :math:`W`, and bias vectors :math:`b`. This function calculates hidden states :math:`h_t` and :math:`c_t` for each time :math:`t` from input :math:`x_t`. .. math:: h_t = f(W_0 x_t + W_1 h_{t-1} + b_0 + b_1) where :math:`f` is an activation function. Weight matrices :math:`W` contains two matrices :math:`W_0` and :math:`W_1`. :math:`W_0` is a parameter for an input sequence. :math:`W_1` is a parameter for a hidden state. Bias matrices :math:`b` contains two matrices :math:`b_0` and :math:`b_1`. :math:`b_0` is a parameter for an input sequence. :math:`b_1` is a parameter for a hidden state. As the function accepts a sequence, it calculates :math:`h_t` for all :math:`t` with one call. Two weight matrices and two bias vectors are required for each layer. So, when :math:`S` layers exist, you need to prepare :math:`2S` weight matrices and :math:`2S` bias vectors. If the number of layers ``n_layers`` is greather than :math:`1`, input of ``k``-th layer is hidden state ``h_t`` of ``k-1``-th layer. Note that all input variables except first layer may have different shape from the first layer. Args: n_layers(int): Number of layers. dropout_ratio(float): Dropout ratio. hx (:class:`~chainer.Variable`): Variable holding stacked hidden states. Its shape is ``(S, B, N)`` where ``S`` is number of layers and is equal to ``n_layers``, ``B`` is mini-batch size, and ``N`` is dimension of hidden units. ws (list of list of :class:`~chainer.Variable`): Weight matrices. ``ws[i]`` represents weights for i-th layer. Each ``ws[i]`` is a list containing two matrices. ``ws[i][j]`` is corresponding with ``W_j`` in the equation. Only ``ws[0][j]`` where ``0 <= j < 1`` is ``(N, I)`` shape as they are multiplied with input variables. All other matrices has ``(N, N)`` shape. bs (list of list of :class:`~chainer.Variable`): Bias vectors. ``bs[i]`` represnents biases for i-th layer. Each ``bs[i]`` is a list containing two vectors. ``bs[i][j]`` is corresponding with ``b_j`` in the equation. Shape of each matrix is ``(N,)`` where ``N`` is dimension of hidden units. xs (list of :class:`~chainer.Variable`): A list of :class:`~chainer.Variable` holding input values. Each element ``xs[t]`` holds input value for time ``t``. Its shape is ``(B_t, I)``, where ``B_t`` is mini-batch size for time ``t``, and ``I`` is size of input units. Note that this function supports variable length sequences. When sequneces has different lengths, sort sequences in descending order by length, and transpose the sorted sequence. :func:`~chainer.functions.transpose_sequence` transpose a list of :func:`~chainer.Variable` holding sequence. So ``xs`` needs to satisfy ``xs[t].shape[0] >= xs[t + 1].shape[0]``. activation (str): Activation function name. Please select ``tanh`` or ``relu``. Returns: tuple: This function returns a tuple containing two elements, ``hy`` and ``ys``. - ``hy`` is an updated hidden states whose shape is same as ``hx``. - ``ys`` is a list of :class:`~chainer.Variable` . Each element ``ys[t]`` holds hidden states of the last layer corresponding to an input ``xs[t]``. Its shape is ``(B_t, N)`` where ``B_t`` is mini-batch size for time ``t``, and ``N`` is size of hidden units. Note that ``B_t`` is the same value as ``xs[t]``. """ return n_step_rnn_base(n_layers, dropout_ratio, hx, ws, bs, xs, activation, use_bi_direction=False, **kwargs) def n_step_birnn( n_layers, dropout_ratio, hx, ws, bs, xs, activation='tanh', **kwargs): """n_step_birnn(n_layers, dropout_ratio, hx, ws, bs, xs, activation='tanh') Stacked Bi-directional RNN function for sequence inputs. This function calculates stacked Bi-directional RNN with sequences. This function gets an initial hidden state :math:`h_0`, an initial cell state :math:`c_0`, an input sequence :math:`x`, weight matrices :math:`W`, and bias vectors :math:`b`. This function calculates hidden states :math:`h_t` and :math:`c_t` for each time :math:`t` from input :math:`x_t`. .. math:: h^{f}_t &=& f(W^{f}_0 x_t + W^{f}_1 h_{t-1} + b^{f}_0 + b^{f}_1), \\\\ h^{b}_t &=& f(W^{b}_0 x_t + W^{b}_1 h_{t-1} + b^{b}_0 + b^{b}_1), \\\\ h_t &=& [h^{f}_t; h^{f}_t], \\\\ where :math:`f` is an activation function. Weight matrices :math:`W` contains two matrices :math:`W^{f}` and :math:`W^{b}`. :math:`W^{f}` is weight matrices for forward directional RNN. :math:`W^{b}` is weight matrices for backward directional RNN. :math:`W^{f}` contains :math:`W^{f}_0` for an input sequence and :math:`W^{f}_1` for a hidden state. :math:`W^{b}` contains :math:`W^{b}_0` for an input sequence and :math:`W^{b}_1` for a hidden state. Bias matrices :math:`b` contains two matrices :math:`b^{f}` and :math:`b^{f}`. :math:`b^{f}` contains :math:`b^{f}_0` for an input sequence and :math:`b^{f}_1` for a hidden state. :math:`b^{b}` contains :math:`b^{b}_0` for an input sequence and :math:`b^{b}_1` for a hidden state. As the function accepts a sequence, it calculates :math:`h_t` for all :math:`t` with one call. Two weight matrices and two bias vectors are required for each layer. So, when :math:`S` layers exist, you need to prepare :math:`2S` weight matrices and :math:`2S` bias vectors. If the number of layers ``n_layers`` is greather than :math:`1`, input of ``k``-th layer is hidden state ``h_t`` of ``k-1``-th layer. Note that all input variables except first layer may have different shape from the first layer. Args: n_layers(int): Number of layers. dropout_ratio(float): Dropout ratio. hx (:class:`~chainer.Variable`): Variable holding stacked hidden states. Its shape is ``(2S, B, N)`` where ``S`` is number of layers and is equal to ``n_layers``, ``B`` is mini-batch size, and ``N`` is dimension of hidden units. Because of bi-direction, the first dimension length is ``2S``. ws (list of list of :class:`~chainer.Variable`): Weight matrices. ``ws[2 * i + di]`` represents weights for i-th layer. Note that ``di = 0`` for forward-RNN and ``di = 1`` for backward-RNN. Each ``ws[2 * i + di]`` is a list containing two matrices. ``ws[2 * i + di][j]`` is corresponding with ``W^{f}_j`` if ``di = 0`` and corresponding with ``W^{b}_j`` if ``di = 1`` in the equation. Only ``ws[0][j]`` and ``ws[1][j]`` where ``0 <= j < 1`` are ``(N, I)`` shape as they are multiplied with input variables. All other matrices has ``(N, N)`` shape. bs (list of list of :class:`~chainer.Variable`): Bias vectors. ``bs[2 * i + di]`` represnents biases for i-th layer. Note that ``di = 0`` for forward-RNN and ``di = 1`` for backward-RNN. Each ``bs[2 * i + di]`` is a list containing two vectors. ``bs[2 * i + di][j]`` is corresponding with ``b^{f}_j`` if ``di = 0`` and corresponding with ``b^{b}_j`` if ``di = 1`` in the equation. Shape of each matrix is ``(N,)`` where ``N`` is dimension of hidden units. xs (list of :class:`~chainer.Variable`): A list of :class:`~chainer.Variable` holding input values. Each element ``xs[t]`` holds input value for time ``t``. Its shape is ``(B_t, I)``, where ``B_t`` is mini-batch size for time ``t``, and ``I`` is size of input units. Note that this function supports variable length sequences. When sequneces has different lengths, sort sequences in descending order by length, and transpose the sorted sequence. :func:`~chainer.functions.transpose_sequence` transpose a list of :func:`~chainer.Variable` holding sequence. So ``xs`` needs to satisfy ``xs[t].shape[0] >= xs[t + 1].shape[0]``. activation (str): Activation function name. Please select ``tanh`` or ``relu``. Returns: tuple: This function returns a tuple containing three elements, ``hy`` and ``ys``. - ``hy`` is an updated hidden states whose shape is same as ``hx``. - ``ys`` is a list of :class:`~chainer.Variable` . Each element ``ys[t]`` holds hidden states of the last layer corresponding to an input ``xs[t]``. Its shape is ``(B_t, N)`` where ``B_t`` is mini-batch size for time ``t``, and ``N`` is size of hidden units. Note that ``B_t`` is the same value as ``xs[t]``. """ return n_step_rnn_base(n_layers, dropout_ratio, hx, ws, bs, xs, activation, use_bi_direction=True) def n_step_rnn_base(n_layers, dropout_ratio, hx, ws, bs, xs, activation, use_bi_direction, **kwargs): """n_step_rnn_base(n_layers, dropout_ratio, hx, ws, bs, xs, activation, \ use_bi_direction) Base function for Stack RNN/BiRNN functions. This function is used at :func:`chainer.functions.n_step_birnn` and :func:`chainer.functions.n_step_rnn`. This function's behavior depends on following arguments, ``activation`` and ``use_bi_direction``. Args: n_layers(int): Number of layers. dropout_ratio(float): Dropout ratio. hx (:class:`~chainer.Variable`): Variable holding stacked hidden states. Its shape is ``(S, B, N)`` where ``S`` is number of layers and is equal to ``n_layers``, ``B`` is mini-batch size, and ``N`` is dimension of hidden units. ws (list of list of :class:`~chainer.Variable`): Weight matrices. ``ws[i]`` represents weights for i-th layer. Each ``ws[i]`` is a list containing two matrices. ``ws[i][j]`` is corresponding with ``W_j`` in the equation. Only ``ws[0][j]`` where ``0 <= j < 1`` is ``(N, I)`` shape as they are multiplied with input variables. All other matrices has ``(N, N)`` shape. bs (list of list of :class:`~chainer.Variable`): Bias vectors. ``bs[i]`` represnents biases for i-th layer. Each ``bs[i]`` is a list containing two vectors. ``bs[i][j]`` is corresponding with ``b_j`` in the equation. Shape of each matrix is ``(N,)`` where ``N`` is dimension of hidden units. xs (list of :class:`~chainer.Variable`): A list of :class:`~chainer.Variable` holding input values. Each element ``xs[t]`` holds input value for time ``t``. Its shape is ``(B_t, I)``, where ``B_t`` is mini-batch size for time ``t``, and ``I`` is size of input units. Note that this function supports variable length sequences. When sequneces has different lengths, sort sequences in descending order by length, and transpose the sorted sequence. :func:`~chainer.functions.transpose_sequence` transpose a list of :func:`~chainer.Variable` holding sequence. So ``xs`` needs to satisfy ``xs[t].shape[0] >= xs[t + 1].shape[0]``. activation (str): Activation function name. Please select ``tanh`` or ``relu``. use_bi_direction (bool): If ``True``, this function uses Bi-directional RNN. Returns: tuple: This function returns a tuple containing three elements, ``hy`` and ``ys``. - ``hy`` is an updated hidden states whose shape is same as ``hx``. - ``ys`` is a list of :class:`~chainer.Variable` . Each element ``ys[t]`` holds hidden states of the last layer corresponding to an input ``xs[t]``. Its shape is ``(B_t, N)`` where ``B_t`` is mini-batch size for time ``t``, and ``N`` is size of hidden units. Note that ``B_t`` is the same value as ``xs[t]``. .. seealso:: :func:`chainer.functions.n_step_rnn` :func:`chainer.functions.n_step_birnn` """ if kwargs: argument.check_unexpected_kwargs( kwargs, train='train argument is not supported anymore. ' 'Use chainer.using_config', use_cudnn='use_cudnn argument is not supported anymore. ' 'Use chainer.using_config') argument.assert_kwargs_empty(kwargs) activation_list = ['tanh', 'relu'] if activation not in activation_list: candidate = ','.join(activation_list) raise ValueError('Invalid activation: "%s". Please select from [%s]' % (activation, candidate)) # Check input size consistency with xs and ws. x_in = xs[0].shape[1] w_in = ws[0][0].shape[1] if x_in != w_in: raise ValueError('Inconsistent input size in input values and weight ' 'parameters: {} != {}'.format(x_in, w_in)) xp = backend.get_array_module(hx) directions = 1 if use_bi_direction: directions = 2 combined = _combine_inputs(hx, ws, bs, xs, n_layers, directions) has_chainerx_array, combined = _extract_apply_in_data(combined) hx_chx, ws_chx, bs_chx, xs_chx = _seperate_inputs( combined, n_layers, len(xs), directions) if has_chainerx_array and xp is chainerx and dropout_ratio == 0: if use_bi_direction: hy, ys = chainerx.n_step_birnn( n_layers, hx_chx, ws_chx, bs_chx, xs_chx, activation) else: hy, ys = chainerx.n_step_rnn( n_layers, hx_chx, ws_chx, bs_chx, xs_chx, activation) hy = variable.Variable._init_unchecked( hy, requires_grad=hy.is_backprop_required(), is_chainerx_array=True) ys = [variable.Variable._init_unchecked( y, requires_grad=y.is_backprop_required(), is_chainerx_array=True) for y in ys] return hy, ys if xp is cuda.cupy and chainer.should_use_cudnn('>=auto', 5000): lengths = [len(x) for x in xs] xs = chainer.functions.concat(xs, axis=0) with chainer.using_device(xs.device): states = cuda.get_cudnn_dropout_states() states.set_dropout_ratio(dropout_ratio) rnn_mode = 'rnn_%s' % activation w = cudnn_rnn_weight_concat( n_layers, states, use_bi_direction, rnn_mode, ws, bs) if use_bi_direction: # Bi-directional RNN if activation == 'tanh': rnn = NStepBiRNNTanh elif activation == 'relu': rnn = NStepBiRNNReLU else: # Uni-directional RNN if activation == 'tanh': rnn = NStepRNNTanh elif activation == 'relu': rnn = NStepRNNReLU hy, ys = rnn(n_layers, states, lengths)(hx, w, xs) sections = numpy.cumsum(lengths[:-1]) ys = chainer.functions.split_axis(ys, sections, 0) return hy, ys else: def f(x, h, c, w, b): xw, hw = w xb, hb = b rnn_in = linear.linear(x, xw, xb) + linear.linear(h, hw, hb) if activation == 'tanh': return tanh.tanh(rnn_in), None elif activation == 'relu': return relu.relu(rnn_in), None hy, _, ys = n_step_rnn_impl( f, n_layers, dropout_ratio, hx, None, ws, bs, xs, use_bi_direction) return hy, ys def n_step_rnn_impl( f, n_layers, dropout_ratio, hx, cx, ws, bs, xs, use_bi_direction): direction = 2 if use_bi_direction else 1 hx = chainer.functions.separate(hx) use_cell = cx is not None if use_cell: cx = chainer.functions.separate(cx) else: cx = [None] * len(hx) xs_next = xs hy = [] cy = [] for layer in six.moves.range(n_layers): # Forward RNN if layer == 0: xs = xs_next else: xs = _dropout_sequence(xs_next, dropout_ratio) idx = direction * layer h, c, h_forward = _one_directional_loop( f, xs, hx[idx], cx[idx], ws[idx], bs[idx]) hy.append(h) cy.append(c) if use_bi_direction: # Backward RNN idx = direction * layer + 1 if layer == 0: xs = xs_next else: xs = _dropout_sequence(xs_next, dropout_ratio) h, c, h_backward = _one_directional_loop( f, reversed(xs), hx[idx], cx[idx], ws[idx], bs[idx]) h_backward.reverse() # Concat xs_next = [concat.concat([hfi, hbi], axis=1) for hfi, hbi in six.moves.zip(h_forward, h_backward)] hy.append(h) cy.append(c) else: # Uni-directional RNN xs_next = h_forward ys = xs_next hy = stack.stack(hy) if use_cell: cy = stack.stack(cy) else: cy = None return hy, cy, tuple(ys) def _one_directional_loop(f, xs, h, c, w, b): h_list = [] for x in xs: batch = len(x) need_split = len(h) > batch if need_split: h, h_rest = split_axis.split_axis(h, [batch], axis=0) if c is not None: c, c_rest = split_axis.split_axis(c, [batch], axis=0) h, c = f(x, h, c, w, b) h_list.append(h) if need_split: h = concat.concat([h, h_rest], axis=0) if c is not None: c = concat.concat([c, c_rest], axis=0) return h, c, h_list def _dropout_sequence(xs, dropout_ratio): return [dropout.dropout(x, ratio=dropout_ratio) for x in xs]
33,371
37.491349
79
py
chainer
chainer-master/chainer/functions/util/forget.py
import chainer from chainer import function from chainer import function_node from chainer import variable def _call_func(func, xs): outs = func(*xs) if isinstance(outs, tuple): for i, out in enumerate(outs): if isinstance(out, variable.Variable): continue n = i + 1 suffix = {1: 'st', 2: 'nd', 3: 'rd'}.get( n if n < 20 else n % 10, 'th') msg = ('{}{} element of a returned tuple is not Variable, ' 'but is {}').format(n, suffix, type(out)) raise RuntimeError(msg) elif isinstance(outs, variable.Variable): outs = (outs,) else: msg = ('A tuple of Variables or a Variable are expected, but {} ' 'is returned.'.format(type(outs))) raise RuntimeError(msg) return outs class Forget(function_node.FunctionNode): def __init__(self, func): if not callable(func): raise TypeError('func must be callable') self.func = func def forward(self, inputs): self.retain_inputs(tuple(range(len(inputs)))) with function.no_backprop_mode(),\ chainer.using_config('_will_recompute', True): xs = [variable.Variable(x) for x in inputs] outs = _call_func(self.func, xs) return tuple(out.data for out in outs) def backward(self, indexes, grad_outputs): # Double backprop is not allowed if chainer.config.enable_backprop: raise RuntimeError('double backpropagation in functions.forget is ' 'not allowed.') inputs = self.get_retained_inputs() # Create new variables that have no creators dummy_inputs = tuple([variable.Variable(inp.array) for inp in inputs]) with function.force_backprop_mode(),\ chainer.using_config('in_recomputing', True): outs = _call_func(self.func, dummy_inputs) assert len(outs) == len(grad_outputs) output_tuples = [] for out, grad_output in zip(outs, grad_outputs): if grad_output is not None: output_tuples.append((out.node, grad_output)) # TODO(kataoka): use outer backward's `retain_grad` and `loss_scale` chainer._backprop._backprop_to_all(output_tuples, False, None) return tuple([inp.grad_var for inp in dummy_inputs]) def forget(func, *xs): """Calls a function without storing intermediate results. On a forward propagation, Chainer normally stores all intermediate results of :class:`~chainer.variable.VariableNode`\\ s on a computational graph as they are required on backward propagation. Sometimes these results consume too much memory. ``F.forget`` *forgets* such intermediate results on forward propagation, and still supports backpropagation with recalculation. On a forward propagation, ``F.forget`` calls a given function with given variables without creating a computational graph. That means, no intermediate results are stored. On a backward propagation, ``F.forget`` calls the given function again to create a computational graph for backpropagation. ``F.forget`` reduces internal memory usage, whereas it requires more calculation time as it calls the function twice. .. admonition:: Example Let ``f`` be a function defined as: >>> def f(a, b): ... return (a + b) * a and, ``x`` and ``y`` be :class:`~chainer.Variable`\\ s: >>> x = chainer.Variable(np.random.uniform(-1, 1, 5).astype(np.float32)) >>> y = chainer.Variable(np.random.uniform(-1, 1, 5).astype(np.float32)) When ``z`` is calculated as ``z = f(x, y)``, its intermediate result ``x + y`` is stored in memory. Instead, if you call ``f`` with ``F.forget``: >>> z = F.forget(f, x, y) intermediate ``x + y`` is forgotten. .. note:: ``F.forget`` does not support functions which behave differently in multiple calls with the same inputs, such as :meth:`F.dropout() <chainer.functions.dropout>` and :meth:`F.negative_sampling() <chainer.functions.negative_sampling>`. .. note:: In case input argument variables are of :ref:`ndarray` objects, arguments will automatically be converted to :class:`~chainer.Variable`\\ s. This conversion takes place to ensure that this function is included in the computational graph to enable backward computations. .. note:: ``F.forget`` does not support double backpropagation. .. note:: If you want to use ``F.forget`` to a link which updates the link's internal information every time the forward computation is called, please ensure that the information is updated just once in a single iteration. You may use the ``chainer.config.in_recomputing`` flag to check if the forward computation is the first call in an iteration. Please see the implementation of :class:`~chainer.links.BatchNormalization` for detail. Args: func (callable): A function to call. It needs to be called with :class:`~chainer.Variable` object(s) and to return a :class:`~chainer.Variable` object or a tuple of :class:`~chainer.Variable` objects. xs (:class:`tuple` of :class:`~chainer.Variable` or :ref:`ndarray`): Argument variables of the function. Returns: ~chainer.Variable: A variable ``func`` returns. If it returns a tuple, the method returns a tuple too. """ xs = tuple(x if isinstance(x, variable.Variable) else variable.Variable(x, requires_grad=True) for x in xs) y = Forget(func).apply(xs) if len(y) == 1: y, = y return y
5,860
36.33121
79
py
chainer
chainer-master/chainer/functions/util/__init__.py
0
0
0
py
chainer
chainer-master/chainer/functions/array/squeeze.py
import six from chainer import backend from chainer import function_node from chainer.utils import type_check def argone(iterable): result = [] for i, x in enumerate(iterable): if not isinstance(x, six.integer_types): raise ValueError('elements in iterable must be int') if x == 1: result.append(i) return result class Squeeze(function_node.FunctionNode): """Remove dimensions of size one from the shape of a ndarray.""" def __init__(self, axis=None): if axis is None: self.axis = None elif isinstance(axis, six.integer_types): self.axis = (axis,) elif isinstance(axis, tuple) and all( isinstance(x, six.integer_types) for x in axis): self.axis = axis else: raise TypeError('axis must be None, int or tuple of ints') def check_type_forward(self, in_types): type_check.expect(in_types.size() == 1) x_type = in_types[0] if self.axis is not None: for x in self.axis: if x >= 0: type_check.expect(x < x_type.ndim) else: type_check.expect(-x_type.ndim <= x) def forward_chainerx(self, inputs): x, = inputs return x.squeeze(self.axis), def forward(self, inputs): x, = inputs xp = backend.get_array_module(x) return xp.squeeze(x, self.axis), def backward(self, indexes, grad_outputs): if self.axis is None: axis = tuple(argone(self.inputs[0].shape)) else: axis = self.axis ndim = len(self.inputs[0].shape) axis = [x + ndim if x < 0 else x for x in axis] axis.sort() gx, = grad_outputs shape = list(gx.shape) for x in axis: # axis needs to be sorted shape.insert(x, 1) return gx.reshape(shape), def squeeze(x, axis=None): """Remove dimensions of size one from the shape of a ndarray. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`(s_1, s_2, ..., s_N)` -shaped float array. axis (None or int or tuple of ints): A subset of the single-dimensional entries in the shape to remove. If ``None`` is supplied, all of them are removed. The dimension index starts at zero. If an axis with dimension greater than one is selected, an error is raised. Returns: ~chainer.Variable: Variable whose dimensions of size 1 are removed. .. admonition:: Example >>> x = np.array([[[[0, 1, 2]]], [[[3, 4, 5]]]], np.float32) >>> x.shape (2, 1, 1, 3) >>> y = F.squeeze(x) >>> y.shape (2, 3) >>> y.array array([[0., 1., 2.], [3., 4., 5.]], dtype=float32) >>> y = F.squeeze(x, axis=1) >>> y.shape (2, 1, 3) >>> y.array array([[[0., 1., 2.]], <BLANKLINE> [[3., 4., 5.]]], dtype=float32) >>> y = F.squeeze(x, axis=(1, 2)) >>> y.shape (2, 3) >>> y.array array([[0., 1., 2.], [3., 4., 5.]], dtype=float32) """ y, = Squeeze(axis).apply((x,)) return y
3,313
28.589286
79
py
chainer
chainer-master/chainer/functions/array/rollaxis.py
import six from chainer import backend from chainer import function_node from chainer.utils import type_check class Rollaxis(function_node.FunctionNode): """Roll axis of an array.""" def __init__(self, axis, start): if not isinstance(axis, six.integer_types): raise TypeError('axis must be int') if not isinstance(start, six.integer_types): raise TypeError('start must be int') self.axis = axis self.start = start def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type = in_types[0] if self.axis >= 0: type_check.expect(x_type.ndim > self.axis) else: type_check.expect(x_type.ndim > -self.axis - 1) if self.start >= 0: type_check.expect(x_type.ndim >= self.start) else: type_check.expect(x_type.ndim > -self.start - 1) def forward(self, inputs): self.retain_inputs(()) self._in_ndim = inputs[0].ndim xp = backend.get_array_module(*inputs) return xp.rollaxis(inputs[0], self.axis, self.start), def backward(self, indexes, gy): axis = self.axis if axis < 0: axis += self._in_ndim start = self.start if start < 0: start += self._in_ndim if axis > start: axis += 1 elif axis < start: start -= 1 return Rollaxis(start, axis).apply(gy) def rollaxis(x, axis, start=0): """Roll the axis backwards to the given position. This function continues to be supported for backward compatibility, but you should prefer ``chainer.functions.moveaxis(x, source, destination)``. See :func:`chainer.functions.moveaxis`. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. axis (int): The axis to roll backwards. start (int): The place to which the axis is moved. Returns: ~chainer.Variable: Variable whose axis is rolled. """ return Rollaxis(axis, start).apply((x,))[0]
2,092
27.283784
73
py
chainer
chainer-master/chainer/functions/array/concat.py
import numpy import six import chainer from chainer import backend from chainer.backends import intel64 from chainer import function_node from chainer.utils import type_check import chainerx class Concat(function_node.FunctionNode): """Concatenate multiple tensors towards specified axis.""" # concat along the channel dimension by default def __init__(self, axis=1): if not isinstance(axis, six.integer_types): raise TypeError('axis must be int') self.axis = axis def check_type_forward(self, in_types): type_check.expect(in_types.size() > 0) type_check.expect(in_types[0].ndim > type_check.make_variable(self.axis, 'axis')) type_check.expect( -in_types[0].ndim <= self.axis, self.axis < in_types[0].ndim ) ndim = type_check.eval(in_types[0].ndim) axis = self.axis % ndim for i in six.moves.range(1, type_check.eval(in_types.size())): type_check.expect( in_types[0].dtype == in_types[i].dtype, in_types[0].ndim == in_types[i].ndim, ) for d in six.moves.range(0, ndim): if d == axis: continue type_check.expect(in_types[0].shape[d] == in_types[i].shape[d]) def forward(self, xs): if (intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(xs, (4,))): # iDeep implementation return self._forward_ideep(xs) # Generic implementation xp = backend.get_array_module(*xs) return xp.concatenate(xs, self.axis), def forward_chainerx(self, xs): return chainerx.concatenate(xs, self.axis), def _forward_ideep(self, xs): xs_mdarray = intel64.ideep.mdarrayVector() for x in xs: xs_mdarray.push_back(intel64.ideep.array(x)) ndim = xs[0].ndim axis = self.axis % ndim return intel64.ideep.concat.Forward(xs_mdarray, axis), def backward(self, indexes, grad_outputs): if len(self.inputs) == 1: return grad_outputs sizes = numpy.array( [v.shape[self.axis] for v in self.inputs[:-1]] ).cumsum() gx, = grad_outputs return chainer.functions.split_axis(gx, sizes, self.axis) def concat(xs, axis=1): """Concatenates given variables along an axis. Args: xs (tuple of :class:`~chainer.Variable` or :ref:`ndarray`): Input variables to be concatenated. The variables must have the \ same shape, except in the dimension corresponding to axis. axis (int): The axis along which the arrays will be joined. Default \ is 1. Returns: ~chainer.Variable: The concatenated variable. .. admonition:: Example >>> x = np.arange(0, 12).reshape(3, 4) >>> x array([[ 0, 1, 2, 3], [ 4, 5, 6, 7], [ 8, 9, 10, 11]]) >>> y = np.arange(0, 3).reshape(3, 1) >>> y array([[0], [1], [2]]) >>> z = F.concat((x, y), axis=1) >>> z.array array([[ 0, 1, 2, 3, 0], [ 4, 5, 6, 7, 1], [ 8, 9, 10, 11, 2]]) """ y, = Concat(axis).apply(xs) return y
3,365
29.6
79
py
chainer
chainer-master/chainer/functions/array/depth2space.py
import numpy import chainer from chainer import backend from chainer import function_node from chainer.utils import type_check class Depth2Space(function_node.FunctionNode): """Depth to space transformation.""" def __init__(self, r): self.r = r def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) type_check.expect( in_types[0].dtype.kind == 'f', in_types[0].ndim == 4 ) def forward(self, inputs): X, = inputs xp = backend.get_array_module(X) bsize, c, a, b = X.shape c //= self.r ** 2 if xp is numpy: # These codes run faster on CPU than below `else` block codes. X = xp.transpose(X, (0, 2, 3, 1)) X = xp.reshape(X, (bsize, a, b, self.r, self.r, c)) X = xp.transpose(X, (0, 1, 3, 2, 4, 5)) X = xp.reshape(X, (bsize, a * self.r, b * self.r, c)) X = xp.transpose(X, (0, 3, 1, 2)) else: X = xp.reshape(X, (bsize, self.r, self.r, c, a, b)) X = xp.transpose(X, (0, 3, 4, 1, 5, 2)) X = xp.reshape(X, (bsize, c, a * self.r, b * self.r)) return X, def backward(self, indexes, grad_outputs): gy, = grad_outputs gy = chainer.functions.space2depth(gy, self.r) return gy, def depth2space(X, r): """Computes the depth2space transformation for subpixel calculations. Args: X (:class:`~chainer.Variable` or :ref:`ndarray`): Variable holding a 4d array of shape ``(batch, channel * r * r, dim1, dim2)``. r (int): the upscaling factor. Returns: ~chainer.Variable: A variable holding the upscaled array from interspersed depth layers. The shape is ``(batch, channel, dim1 * r, dim2 * r)``. .. note:: This can be used to compute super-resolution transformations. See https://arxiv.org/abs/1609.05158 for details. .. seealso:: :func:`space2depth` .. admonition:: Example >>> X = np.arange(24).reshape(1, 4, 2, 3).astype(np.float32) >>> X.shape (1, 4, 2, 3) >>> X array([[[[ 0., 1., 2.], [ 3., 4., 5.]], <BLANKLINE> [[ 6., 7., 8.], [ 9., 10., 11.]], <BLANKLINE> [[12., 13., 14.], [15., 16., 17.]], <BLANKLINE> [[18., 19., 20.], [21., 22., 23.]]]], dtype=float32) >>> y = F.depth2space(X, 2) >>> y.shape (1, 1, 4, 6) >>> y.array array([[[[ 0., 6., 1., 7., 2., 8.], [12., 18., 13., 19., 14., 20.], [ 3., 9., 4., 10., 5., 11.], [15., 21., 16., 22., 17., 23.]]]], dtype=float32) """ return Depth2Space(r).apply((X,))[0]
2,907
29.291667
76
py
chainer
chainer-master/chainer/functions/array/permutate.py
import numpy import six import chainer from chainer import backend from chainer.backends import cuda from chainer import function_node from chainer.utils import type_check def _check_indices(indices): if len(indices) == 0: return # TODO(unno): Check indices without cpu indices = cuda.to_cpu(indices) for i in indices: if 0 <= i < len(indices): continue raise ValueError('Out of bounds index: {}'.format(i)) sort = numpy.sort(indices) for s, t in six.moves.zip(sort, sort[1:]): if s == t: raise ValueError('indices contains duplicate value: {}'.format(s)) def _inverse_indices(indices): xp = backend.get_array_module(indices) r = xp.empty_like(indices) if xp is numpy: r[indices] = numpy.arange(len(indices)) else: cuda.elementwise( 'S ind', 'raw S r', 'r[ind] = i', 'inverse_indices' )(indices, r) return r class Permutate(function_node.FunctionNode): """Permutate function.""" def __init__(self, indices, axis, inv): self.indices = indices self.axis = axis self.inv = inv def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type, = in_types if self.axis < 0: type_check.expect(x_type.ndim >= -self.axis) else: type_check.expect(x_type.ndim > self.axis) def _permutate(self, x, indices, inv): if inv: indices = _inverse_indices(indices) return x[((slice(None),) * self.axis) + (indices,)] def forward(self, inputs): x, = inputs inds = self.indices if chainer.is_debug(): _check_indices(inds) return self._permutate(x, inds, self.inv), def backward(self, indexes, grad_outputs): g, = grad_outputs inds = self.indices gx, = Permutate(inds, self.axis, not self.inv).apply((g,)) return gx, def permutate(x, indices, axis=0, inv=False): """Permutates a given variable along an axis. This function permutate ``x`` with given ``indices``. That means ``y[i] = x[indices[i]]`` for all ``i``. Note that this result is same as ``y = x.take(indices)``. ``indices`` must be a permutation of ``[0, 1, ..., len(x) - 1]``. When ``inv`` is ``True``, ``indices`` is treated as its inverse. That means ``y[indices[i]] = x[i]``. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Variable to permutate. A :math:`(s_1, s_2, ..., s_N)` -shaped float array. indices (:class:`~chainer.Variable` or :ref:`ndarray`): Indices to extract from the variable. A one-dimensional int array. axis (int): Axis that the input array is permutate along. inv (bool): If ``True``, ``indices`` is treated as its inverse. Returns: ~chainer.Variable: Output variable. .. admonition:: Example >>> x = np.arange(6).reshape((3, 2)).astype(np.float32) >>> x array([[0., 1.], [2., 3.], [4., 5.]], dtype=float32) >>> indices = np.array([2, 0, 1], np.int32) >>> y = F.permutate(x, indices) >>> y.array array([[4., 5.], [0., 1.], [2., 3.]], dtype=float32) >>> y = F.permutate(x, indices, inv=True) >>> y.array array([[2., 3.], [4., 5.], [0., 1.]], dtype=float32) >>> indices = np.array([1, 0], np.int32) >>> y = F.permutate(x, indices, axis=1) >>> y.array array([[1., 0.], [3., 2.], [5., 4.]], dtype=float32) """ if indices.dtype.kind != 'i' or indices.ndim != 1: raise ValueError( 'indices should be a one-dimensional int array') if isinstance(indices, chainer.Variable): indices = indices.array y, = Permutate(indices, axis, inv).apply((x,)) return y
4,025
28.822222
78
py
chainer
chainer-master/chainer/functions/array/separate.py
from chainer import backend from chainer import function_node from chainer.functions.array import stack from chainer.utils import type_check class Separate(function_node.FunctionNode): """Function that separates a given array.""" def __init__(self, axis): self.axis = axis def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type = in_types[0] if self.axis >= 0: type_check.expect(self.axis < x_type.ndim) else: type_check.expect(-self.axis <= x_type.ndim) def forward(self, inputs): x, = inputs self._xp = backend.get_array_module(x) xs = self._xp.split(x, x.shape[self.axis], self.axis) ys = [self._xp.squeeze(y, self.axis) for y in xs] self._shape = ys[0].shape self._dtype = x.dtype return tuple(ys) def backward(self, indexes, grad_outputs): grad_outputs = [ self._xp.zeros(self._shape, dtype=self._dtype) if g is None else g for g in grad_outputs] return stack.stack(grad_outputs, self.axis), def separate(x, axis=0): """Separates an array along a given axis. This function separates an array along a given axis. For example, shape of an array is ``(2, 3, 4)``. When it separates the array with ``axis=1``, it returns three ``(2, 4)`` arrays. This function is an inverse of :func:`chainer.functions.stack`. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Variable to be separated. A :math:`(s_1, s_2, ..., s_N)` -shaped float array. axis (int): Axis along which variables are separated. Returns: tuple of chainer.Variable: Output variables. .. seealso:: :func:`chainer.functions.stack` .. admonition:: Example >>> x = np.arange(6).reshape((2, 3)).astype(np.float32) >>> x array([[0., 1., 2.], [3., 4., 5.]], dtype=float32) >>> x.shape (2, 3) >>> y = F.separate(x) # split along axis=0 >>> isinstance(y, tuple) True >>> len(y) 2 >>> y[0].shape (3,) >>> y[0].array array([0., 1., 2.], dtype=float32) >>> y = F.separate(x, axis=1) >>> len(y) 3 >>> y[0].shape (2,) >>> y[0].array array([0., 3.], dtype=float32) """ return Separate(axis).apply((x,))
2,459
27.941176
78
py
chainer
chainer-master/chainer/functions/array/copy.py
import chainer from chainer import backend from chainer.backends import cuda from chainer import function_node from chainer.utils import type_check import chainerx class Copy(function_node.FunctionNode): """Copies the input variable onto the specified device.""" def __init__(self, in_device, out_device): self._in_device = in_device self.out_device = out_device def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) def forward(self, inputs): x, = inputs return self.out_device.send(x), def forward_chainerx(self, inputs): x, = inputs return x.to_device(self.out_device.device), def backward(self, indexes, grad_outputs): f = Copy(self.out_device, self._in_device) return f.apply(grad_outputs) # TODO(niboshi): Link from `dst` to an appropriate device specifier docs. def copy(x, dst): """Copies the input variable onto the specified device. If the input ``x`` already resides on the device specified by ``dst``, no copy will actually take place and the returned variable will hold a view of the input. In other cases, the input will be copied to ``dst``. When ``dst == -1``, the array is copied to the host memory. This function supports copies from host to host, from host to device, from device to device and from device to host. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Variable to be copied. dst: Target device specifier. Returns: ~chainer.Variable: Output variable. .. admonition:: Example >>> import chainer.backends.cuda as cuda >>> x_arr = np.random.uniform(-1, 1, (5, 10)) >>> x = chainer.Variable(x_arr) >>> x.device <CpuDevice (numpy)> >>> y = F.copy(x, '@cupy:0') # from CPU (NumPy) to GPU 0 (CuPy) >>> y.device <GpuDevice (cupy):0> .. note:: Copies between non-ChainerX devices and ChainerX devices are not supported. """ # For backward compatibility if dst is cuda.DummyDevice: dst = chainer.get_device('@numpy') x_is_var = isinstance(x, chainer.Variable) in_device = backend.get_device_from_array(x.array if x_is_var else x) out_device = chainer.get_device(dst) if in_device.xp is chainerx: x_arr = x.chx_array if x_is_var else x if out_device.xp is not chainerx: # ChainerX to non-ChainerX if x_arr.is_backprop_required(): raise RuntimeError( 'F.copy does not support copy from a ChainerX array with ' 'backprop required to a non-ChainerX device.\n' 'From: {}\n' 'To: {}'.format(in_device, out_device)) return chainer.Variable( out_device.send(x_arr), requires_grad=False) # ChainerX to ChainerX return chainer.Variable( out_device.send(x_arr), requires_grad=x_arr.is_backprop_required()) if out_device.xp is chainerx: # Non-ChainerX to ChainerX if x_is_var and x.requires_grad: raise RuntimeError( 'F.copy does not support copy from a non-ChainerX array with ' 'backprop required to a ChainerX device.\n' 'From: {}\n' 'To: {}'.format(in_device, out_device)) x_arr = x.array if x_is_var else x return chainer.Variable(out_device.send(x_arr), requires_grad=False) # Non-ChainerX to non-ChainerX y, = Copy(in_device, out_device).apply((x,)) return y
3,635
32.981308
79
py
chainer
chainer-master/chainer/functions/array/get_item.py
import numpy import chainer from chainer import backend from chainer import function_node from chainer import utils from chainer.utils import type_check from chainer import variable import chainerx _numpy_supports_0d_bool_index = \ numpy.lib.NumpyVersion(numpy.__version__) >= '1.13.0' class GetItem(function_node.FunctionNode): """Function that slices array and extract elements.""" def __init__(self, slices): if isinstance(slices, list): if all([isinstance(s, int) for s in slices]): slices = slices, slices = tuple(slices) elif not isinstance(slices, tuple): slices = slices, if chainer.is_debug(): n_ellipses = 0 for s in slices: if s is Ellipsis: n_ellipses += 1 if n_ellipses > 1: raise ValueError('Only one Ellipsis is allowed') self.slices = slices def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) def forward(self, xs): slices = tuple([ backend.from_chx(s) if isinstance(s, chainerx.ndarray) else s for s in self.slices]) return utils.force_array(xs[0][slices]), def backward(self, indexes, gy): return GetItemGrad( self.slices, self.inputs[0].shape).apply(gy) class GetItemGrad(function_node.FunctionNode): def __init__(self, slices, in_shape): self.slices = slices self._in_shape = in_shape def forward(self, inputs): slices = tuple([ backend.from_chx(s) if isinstance(s, chainerx.ndarray) else s for s in self.slices]) gy, = inputs xp = backend.get_array_module(*inputs) gx = xp.zeros(self._in_shape, gy.dtype) if xp is numpy: try: numpy.add.at(gx, slices, gy) except IndexError: done = False # In numpy<1.13, 0-dim boolean index is not supported in # numpy.add.at and it's supported for 0-dim arr in # arr.__getitem__. if not _numpy_supports_0d_bool_index and len(slices) == 1: idx = numpy.asanyarray(slices[0]) if idx.dtype == numpy.dtype(bool): # Convert the array and the mask to 1-dim. # numpy.add.at with them is supported in older numpy. numpy.add.at(gx[None], idx[None], gy) done = True if not done: msg = ''' GetItem does not support backward for this slices. The slices argument is not supported by numpy.add.at, while it is supported by numpy.ndarray.__getitem__. Please report this error to the issue tracker with the stack trace, the information of your environment, and your script: https://github.com/chainer/chainer/issues/new. ''' raise IndexError(msg) else: gx.scatter_add(slices, inputs[0]) return gx, def backward(self, indexes, ggx): return GetItem(self.slices).apply(ggx) def get_item(x, slices): """Extract elements from array with specified shape, axes and offsets. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): A variable to be sliced. slices (int, slice, Ellipsis, None, integer array-like, boolean\ array-like or tuple of them): An object to specify the selection of elements. Returns: A :class:`~chainer.Variable` object which contains sliced array of ``x``. .. note:: It only supports types that are supported by CUDA's atomicAdd when an integer array is included in ``slices``. The supported types are ``numpy.float32``, ``numpy.int32``, ``numpy.uint32``, ``numpy.uint64`` and ``numpy.ulonglong``. .. note:: It does not support ``slices`` that contains multiple boolean arrays. .. note:: See NumPy documentation for details of `indexing <https://docs.scipy.org/doc/numpy/reference/arrays.indexing.html>`_. .. admonition:: Example >>> x = np.arange(12).reshape((2, 2, 3)) >>> x array([[[ 0, 1, 2], [ 3, 4, 5]], <BLANKLINE> [[ 6, 7, 8], [ 9, 10, 11]]]) >>> F.get_item(x, 0) variable([[0, 1, 2], [3, 4, 5]]) >>> F.get_item(x, (0, 0, slice(0, 2, 1))) # equals x[0, 0, 0:2:1] variable([0, 1]) >>> F.get_item(x, (Ellipsis, 2)) # equals x[..., 2] variable([[ 2, 5], [ 8, 11]]) >>> F.get_item(x, (1, np.newaxis, 1, 0)) # equals x[1, None, 1, 0] variable([9]) """ return GetItem(slices).apply((x,))[0] def install_variable_get_item(): variable.Variable.__getitem__ = get_item
4,935
30.641026
78
py
chainer
chainer-master/chainer/functions/array/flipud.py
from chainer import backend from chainer import function_node from chainer.utils import type_check class FlipUD(function_node.FunctionNode): """Flip array in the up/down direction.""" def check_type_forward(self, in_types): type_check._argname(in_types, ('a',)) a_type = in_types[0] type_check.expect( a_type.dtype.kind == 'f', a_type.ndim >= 1 ) def forward(self, inputs): xp = backend.get_array_module(*inputs) return xp.flipud(inputs[0]), def backward(self, indexes, grad_outputs): return FlipUD().apply(grad_outputs) def flipud(a): """Flip array in the up/down direction. Args: a (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. Returns: ~chainer.Variable: Output variable. """ return FlipUD().apply((a,))[0]
872
21.973684
73
py
chainer
chainer-master/chainer/functions/array/spatial_transformer_sampler.py
import numpy import chainer from chainer import backend from chainer.backends import cuda from chainer import function from chainer.utils import argument from chainer.utils import type_check if cuda.cudnn_enabled: cudnn = cuda.cudnn libcudnn = cuda.libcudnn _sampler_type = cuda.libcudnn.CUDNN_SAMPLER_BILINEAR class SpatialTransformerSampler(function.Function): def check_type_forward(self, in_types): n_in = in_types.size() type_check.expect(2 == n_in) x_type = in_types[0] grid_type = in_types[1] type_check.expect( x_type.dtype.kind == 'f', grid_type.dtype == x_type.dtype, x_type.ndim == 4, grid_type.ndim == 4, grid_type.shape[1] == 2, x_type.shape[0] == grid_type.shape[0], ) def forward_cpu(self, inputs): return self._forward(inputs) def forward_gpu(self, inputs): if not chainer.should_use_cudnn('>=auto', 5000): return self._forward(inputs) x, grid = inputs out_shape = x.shape[:2] + grid.shape[2:] y = cuda.cupy.empty(out_shape, dtype=x.dtype) shape = numpy.array(out_shape, dtype=numpy.int32) x = cuda.cupy.ascontiguousarray(x) grid_t = cuda.cupy.transpose(grid, (0, 2, 3, 1)) grid_t = cuda.cupy.ascontiguousarray(grid_t) handle = cudnn.get_handle() x_desc = cudnn.create_tensor_descriptor(x) y_desc = cudnn.create_tensor_descriptor(y) self.st_desc =\ cuda.cupy.cudnn.create_spatial_transformer_descriptor( _sampler_type, grid.dtype, len(shape), shape.ctypes.data) dtype = numpy.float64 if x.dtype == numpy.float64 else numpy.float32 one = numpy.array(1, dtype=dtype).ctypes zero = numpy.array(0, dtype=dtype).ctypes libcudnn.spatialTfSamplerForward( handle, self.st_desc.value, one.data, x_desc.value, x.data.ptr, grid_t.data.ptr, zero.data, y_desc.value, y.data.ptr) return y, def _forward(self, inputs): x, grid = inputs xp = backend.get_array_module(x) B, C, H, W = x.shape _, _, out_H, out_W = grid.shape grid = grid.reshape(grid.shape[:2] + (-1,)) u = grid[:, 0] v = grid[:, 1] # Pad the image so that pixels locating outside of the original # image's size can be sampled. x_pad = xp.pad(x, ((0, 0), (0, 0), (1, 1), (1, 1)), mode='constant') # Rescale coordinates from [-1, 1] to [0, width or height - 1], # and adjust them to the padded image. u = (u + 1) * (W - 1) / 2 + 1 v = (v + 1) * (H - 1) / 2 + 1 u_clipped = u.clip(0, W + 1) v_clipped = v.clip(0, H + 1) # indices of the 2x2 pixel neighborhood surrounding the coordinates u0 = xp.floor(u_clipped).astype(numpy.int32) u0 = u0.clip(0, W) u1 = u0 + 1 v0 = xp.floor(v_clipped).astype(numpy.int32) v0 = v0.clip(0, H) v1 = v0 + 1 # weights w1 = (u1 - u_clipped) * (v1 - v_clipped) w2 = (u_clipped - u0) * (v1 - v_clipped) w3 = (u1 - u_clipped) * (v_clipped - v0) w4 = (u_clipped - u0) * (v_clipped - v0) w1 = w1.astype(x_pad.dtype, copy=False) w2 = w2.astype(x_pad.dtype, copy=False) w3 = w3.astype(x_pad.dtype, copy=False) w4 = w4.astype(x_pad.dtype, copy=False) x_indexed_1 = xp.concatenate([xp.expand_dims( x_pad[b, :, v0[b], u0[b]], axis=0) for b in range(B)], axis=0) x_indexed_2 = xp.concatenate([xp.expand_dims( x_pad[b, :, v0[b], u1[b]], axis=0) for b in range(B)], axis=0) x_indexed_3 = xp.concatenate([xp.expand_dims( x_pad[b, :, v1[b], u0[b]], axis=0) for b in range(B)], axis=0) x_indexed_4 = xp.concatenate([xp.expand_dims( x_pad[b, :, v1[b], u1[b]], axis=0) for b in range(B)], axis=0) y = w1[:, :, None] * x_indexed_1 y += w2[:, :, None] * x_indexed_2 y += w3[:, :, None] * x_indexed_3 y += w4[:, :, None] * x_indexed_4 y = y.reshape(B, out_H, out_W, C).transpose(0, 3, 1, 2) return y, def backward_cpu(self, inputs, grad_outputs): return self._backward(inputs, grad_outputs) def backward_gpu(self, inputs, grad_outputs): if not chainer.should_use_cudnn('>=auto', 5000): return self._backward(inputs, grad_outputs) x, grid = inputs gy, = grad_outputs grid_t = cuda.cupy.transpose(grid, (0, 2, 3, 1)) grid_t = cuda.cupy.ascontiguousarray(grid_t) x = cuda.cupy.ascontiguousarray(x) gy = cuda.cupy.ascontiguousarray(gy) gx = cuda.cupy.empty_like(x) ggrid_t = cuda.cupy.empty_like(grid_t) handle = cudnn.get_handle() x_desc = cudnn.create_tensor_descriptor(x) dx_desc = cudnn.create_tensor_descriptor(gx) dy_desc = cudnn.create_tensor_descriptor(gy) dtype = numpy.float64 if x.dtype == numpy.float64 else numpy.float32 one = numpy.array(1, dtype=dtype).ctypes zero = numpy.array(0, dtype=dtype).ctypes libcudnn.spatialTfSamplerBackward( handle, self.st_desc.value, one.data, x_desc.value, x.data.ptr, zero.data, dx_desc.value, gx.data.ptr, one.data, dy_desc.value, gy.data.ptr, grid_t.data.ptr, zero.data, ggrid_t.data.ptr) ggrid = cuda.cupy.transpose(ggrid_t, axes=(0, 3, 1, 2)) return gx, ggrid def _backward(self, inputs, grad_outputs): x, grid = inputs xp = backend.get_array_module(x) gy, = grad_outputs B, C, H, W = x.shape _, _, out_H, out_W = grid.shape grid = grid.reshape(grid.shape[:2] + (-1,)) u = grid[:, 0] v = grid[:, 1] # Pad the image so that points locating outside of the original # image's size can be sampled. x_pad = xp.pad(x, ((0, 0), (0, 0), (1, 1), (1, 1)), mode='constant') # Rescale coordinates from [-1, 1] to [0, width or height - 1], # and adjust them to the padded image. u = (u + 1) * (W - 1) / 2 + 1 v = (v + 1) * (H - 1) / 2 + 1 u_clipped = u.clip(0, W + 1) v_clipped = v.clip(0, H + 1) # indices of the 2x2 pixel neighborhood surrounding the coordinates u0 = xp.floor(u_clipped).astype(numpy.int32) u0 = u0.clip(0, W) u1 = u0 + 1 v0 = xp.floor(v_clipped).astype(numpy.int32) v0 = v0.clip(0, H) v1 = v0 + 1 # weights wu0 = u_clipped - u0 wu1 = u1 - u_clipped wv0 = v_clipped - v0 wv1 = v1 - v_clipped wu0 = wu0.astype(gy.dtype, copy=False) wu1 = wu1.astype(gy.dtype, copy=False) wv0 = wv0.astype(gy.dtype, copy=False) wv1 = wv1.astype(gy.dtype, copy=False) # --- gu, gv x_indexed_1 = xp.concatenate([xp.expand_dims( x_pad[b, :, v0[b], u0[b]], axis=0) for b in range(B)], axis=0) x_indexed_2 = xp.concatenate([xp.expand_dims( x_pad[b, :, v0[b], u1[b]], axis=0) for b in range(B)], axis=0) x_indexed_3 = xp.concatenate([xp.expand_dims( x_pad[b, :, v1[b], u0[b]], axis=0) for b in range(B)], axis=0) x_indexed_4 = xp.concatenate([xp.expand_dims( x_pad[b, :, v1[b], u1[b]], axis=0) for b in range(B)], axis=0) gu = -wv1[:, :, None] * x_indexed_1 gu += wv1[:, :, None] * x_indexed_2 gu -= wv0[:, :, None] * x_indexed_3 gu += wv0[:, :, None] * x_indexed_4 gv = -wu1[:, :, None] * x_indexed_1 gv -= wu0[:, :, None] * x_indexed_2 gv += wu1[:, :, None] * x_indexed_3 gv += wu0[:, :, None] * x_indexed_4 gu = gu.reshape(B, out_H, out_W, C).transpose(0, 3, 1, 2) gv = gv.reshape(B, out_H, out_W, C).transpose(0, 3, 1, 2) gu *= gy gv *= gy gu = xp.sum(gu, axis=1) gv = xp.sum(gv, axis=1) # Offsets scaling of the coordinates and clip gradients. u_reshaped = u.reshape(gu.shape) v_reshaped = v.reshape(gv.shape) gu = gu / 2. * (W - 1) * (u_reshaped > 0) * (u_reshaped < (W + 1)) gv = gv / 2. * (H - 1) * (v_reshaped > 0) * (v_reshaped < (H + 1)) ggrid = xp.concatenate((gu[:, None], gv[:, None]), axis=1) # --- gx if xp is numpy: scatter_add = numpy.add.at else: scatter_add = cuda.cupyx.scatter_add gx = xp.zeros_like(x_pad) gy = gy.reshape(B, C, -1) for b in range(B): scatter_add(gx[b], (slice(None), v0[b], u0[b]), gy[b] * wu1[b] * wv1[b]) scatter_add(gx[b], (slice(None), v0[b], u1[b]), gy[b] * wu0[b] * wv1[b]) scatter_add(gx[b], (slice(None), v1[b], u0[b]), gy[b] * wu1[b] * wv0[b]) scatter_add(gx[b], (slice(None), v1[b], u1[b]), gy[b] * wu0[b] * wv0[b]) gx = gx[:, :, 1:-1, 1:-1] return gx, ggrid def spatial_transformer_sampler(x, grid, **kwargs): """2D Spatial Transformer sampler. This is a differentiable image sampler. With a set of sampling points ``grid`` and an input feature map ``x``, this produces a sampled output feature map. This function currently only supports bilinear interpolation as a sampling kernel. When coordinates in ``grid`` is outside range :math:`[-1, 1]`, values are sampled from a zero padded input image. Notation: here is a notation for dimensionalities. - :math:`n` is the batch size. - :math:`c_I` is the number of the input channels. - :math:`h` and :math:`w` are the height and width of the input image, respectively. - :math:`h_O` and :math:`w_O` are the height and width of the output image. See detail in the following paper: `Spatial Transformer Networks <https://arxiv.org/abs/1506.02025>`_. .. note:: cuDNN supports SpatialTransformerSampler from version 5.0.0. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable of shape :math:`(n, c_I, h, w)`. grid (~chainer.Variable): Coordinate variable of shape :math:`(n, 2, h_O, w_O)`. Each coordinate defines the spatial location in the input where a sampling kernel is applied to get the value at a particular pixel in the output. ``grid[idx, :, i, j]`` corresponds to the coordinate that is used to sample the values for an output pixel at location :math:`(i, j)`. In the second dimension, the first coordinate corresponds to the location along the horizontal axis, and the second coordinate corresponds to the location along the vertical axis. The coordinate :math:`(-1, -1)` corresponds to the upper-left corner of the input image. Returns: ~chainer.Variable: Output feature map of shape \ :math:`(n, c_I, h_O, w_O)`. """ if kwargs: argument.check_unexpected_kwargs( kwargs, use_cudnn='The argument "use_cudnn" is not ' 'supported anymore. ' 'Use chainer.using_config(\'use_cudnn\', value) ' 'context where value can be `always`, `never`, or `auto`.') argument.assert_kwargs_empty(kwargs) return SpatialTransformerSampler()(x, grid)
11,635
35.939683
78
py
chainer
chainer-master/chainer/functions/array/as_strided.py
import numpy as np import six from chainer import backend from chainer.backends import cuda from chainer import function_node from chainer.utils import type_check index_dtype = {t().itemsize: t for t in np.sctypes['int']} def _byte2step(iterable, itemsize): for i in iterable: assert i % itemsize == 0 return tuple([i // itemsize for i in iterable]) def _step2byte(iterable, itemsize): return tuple([i * itemsize for i in iterable]) def _maybe_overlapping_memory(shape, strides): """Returns bool value indicating the array with such shape and strides might have overlapping memory. Args: shape (tuple of int): The shape of output. strides (tuple of int): The strides of output, given in the unit of steps. storage_offset (int): The offset between the head of allocated memory and the pointer of first element, given in the unit of steps. Returns: bool: Existence of the overlapping memory """ max_ptr_in_slice = 0 for stride, size in sorted(zip([abs(s) for s in strides], shape)): if stride <= max_ptr_in_slice: return True max_ptr_in_slice += stride * (size - 1) return False def _min_index(shape, strides, storage_offset): """Returns the leftest index in the array (in the unit-steps) Args: shape (tuple of int): The shape of output. strides (tuple of int): The strides of output, given in the unit of steps. storage_offset (int): The offset between the head of allocated memory and the pointer of first element, given in the unit of steps. Returns: int: The leftest pointer in the array """ sh_st_neg = [sh_st for sh_st in zip(shape, strides) if sh_st[1] < 0] if not sh_st_neg: return storage_offset else: return storage_offset + six.moves.reduce( lambda base, sh_st: base + (sh_st[0] - 1) * sh_st[1], sh_st_neg, 0) def _max_index(shape, strides, storage_offset): """Returns the rightest index in the array Args: shape (tuple of int): The shape of output. strides (tuple of int): The strides of output, given in unit-steps. storage_offset (int): The offset between the head of allocated memory and the pointer of first element, given in the unit of steps. Returns: int: The rightest pointer in the array """ sh_st_pos = [sh_st for sh_st in zip(shape, strides) if sh_st[1] > 0] if not sh_st_pos: return storage_offset else: return storage_offset + six.moves.reduce( lambda base, sh_st: base + (sh_st[0] - 1) * sh_st[1], sh_st_pos, 0) def _index_add(augend, indices, addend): """Wrapper of :func:`cupyx.scatter_add` and :func:`numpy.add.at` Args: augend (:class:`numpy.ndarray` or :class:`cupy.ndarray`): The array modified in-place. indices (:class:`numpy.ndarray` or :class:`cupy.ndarray`): The indices of ``augend``. The shape is the same to the ``addend``. addend (:class:`numpy.ndarray` or :class:`cupy.ndarray`): The array to be added. Returns: None """ if isinstance(augend, cuda.ndarray): cuda.cupyx.scatter_add(augend, indices, addend) elif isinstance(augend, np.ndarray): np.add.at(augend, indices, addend) def _get_base_array(array): """Get the founder of :class:`numpy.ndarray`. Args: array (:class:`numpy.ndarray`): The view of the base array. Returns: :class:`numpy.ndarray`: The base array. """ base_array_candidate = array while base_array_candidate.base is not None: base_array_candidate = base_array_candidate.base return base_array_candidate def _stride_array(array, shape, strides, storage_offset): """Wrapper of :func:`numpy.lib.stride_tricks.as_strided`. .. note: ``strides`` and ``storage_offset`` is given in the unit of steps instead the unit of bytes. This specification differs from that of :func:`numpy.lib.stride_tricks.as_strided`. Args: array (:class:`numpy.ndarray` of :class:`cupy.ndarray`): The base array for the returned view. shape (tuple of int): The shape of the returned view. strides (tuple of int): The strides of the returned view, given in the unit of steps. storage_offset (int): The offset from the leftest pointer of allocated memory to the first element of returned view, given in the unit of steps. Returns: :class:`numpy.ndarray` or :class:`cupy.ndarray`: The new view for the base array. """ min_index = _min_index(shape, strides, storage_offset) max_index = _max_index(shape, strides, storage_offset) strides = _step2byte(strides, array.itemsize) storage_offset, = _step2byte((storage_offset,), array.itemsize) if min_index < 0: raise ValueError('Out of buffer: too small index was specified') if isinstance(array, cuda.ndarray): pooled_memory = array.data.mem if (max_index + 1) * array.itemsize > pooled_memory.size: raise ValueError('Out of buffer: too large index was specified') memptr = cuda.cupy.cuda.memory.MemoryPointer(pooled_memory, storage_offset) return cuda.cupy.ndarray(shape, array.dtype, memptr, strides) elif isinstance(array, np.ndarray): base_array = _get_base_array(array) if (max_index + 1) * base_array.itemsize > base_array.nbytes: raise ValueError('Out of buffer: too large index was specified') return np.ndarray(shape, base_array.dtype, base_array.data, storage_offset, strides) else: raise TypeError('Only (np|cp).ndarray is accepted') class TensorGeometry(object): def __init__(self, array): self.shape = array.shape self.strides = _byte2step(array.strides, array.itemsize) if isinstance(array, np.ndarray): base_array = _get_base_array(array) array_ptr = array.__array_interface__['data'][0] base_array_ptr = base_array.__array_interface__['data'][0] offset_bytes = array_ptr - base_array_ptr elif isinstance(array, cuda.ndarray): offset_bytes = array.data.ptr - array.data.mem.ptr else: raise ValueError('only (np|cp).ndarray is supported') self.storage_offset, = _byte2step((offset_bytes,), array.itemsize) self.itemsize = array.itemsize @property def ndim(self): return len(self.shape) class AsStrided(function_node.FunctionNode): """Transportation of :func:`torch.Tensor.as_strided`. While :func:`torch.Tensor.as_strided` does not support nagative strides, this implementation does support it. """ def __init__(self, shape, strides, storage_offset=None): self.shape = shape self.strides = strides self.storage_offset = storage_offset self.input_geometry = None def check_type_forward(self, in_types): type_check.expect(in_types.size() == 1) def forward(self, inputs): assert len(inputs) > 0 x = inputs[0] self.input_geometry = TensorGeometry(x) if self.storage_offset is None: self.storage_offset = self.input_geometry.storage_offset return _stride_array(x, self.shape, self.strides, self.storage_offset), def backward(self, _, grad_outputs): """Backward computation which calls :class:`AsStridedGrad`. .. note: While this implementation is based on *New-Style Function Implementation*, the backward computation does not support double-backpropagation due to *layout agnostic* algorithm ( originally named in the note of pytorch). """ return AsStridedGrad(self.input_geometry, self.shape, self.strides, self.storage_offset).apply(grad_outputs) class AsStridedGrad(function_node.FunctionNode): """Backward of :func:`~chainer.functions.as_strided`. """ def __init__(self, input_geometry, shape, strides, storage_offset): self.input_geometry = input_geometry self.shape = shape self.strides = strides self.storage_offset = storage_offset def forward(self, grads): assert len(grads) > 0 gy = grads[0] if gy.dtype not in np.sctypes['float']: raise TypeError('Only float is supported for back propagation') xp = backend.get_array_module(gy) input_geometry = self.input_geometry itemsize = input_geometry.itemsize if 0 in input_geometry.shape: return xp.zeros(input_geometry.shape) # 1. remove redundant axis from input/output # [redundant axis] # axis with shape==0, shape==1 or strides==0 if 0 in gy.shape: return backend.get_array_module(gy).zeros(input_geometry.shape) else: out_shape = tuple([ self.shape[i] for i in six.moves.range(gy.ndim) if self.shape[i] != 1 and self.strides[i] != 0]) out_strides = tuple([ self.strides[i] for i in six.moves.range(gy.ndim) if self.shape[i] != 1 and self.strides[i] != 0]) gy = gy.sum( tuple([i for i in six.moves.range(gy.ndim) if self.strides[i] == 0])) gy = gy.squeeze() out_storage_offset = self.storage_offset inp_shape = tuple([input_geometry.shape[i] for i in six.moves.range(input_geometry.ndim) if input_geometry.shape[i] != 1]) inp_strides = tuple([input_geometry.strides[i] for i in six.moves.range(input_geometry.ndim) if input_geometry.shape[i] != 1]) inp_storage_offset = input_geometry.storage_offset # 2. calculate minimum required storage for gradient computation inp_min_ptr = _min_index(inp_shape, inp_strides, input_geometry.storage_offset) out_min_ptr = _min_index(out_shape, out_strides, self.storage_offset) common_min_ptr = min(inp_min_ptr, out_min_ptr) inp_max_ptr = _max_index(inp_shape, inp_strides, input_geometry.storage_offset) out_max_ptr = _max_index(out_shape, out_strides, self.storage_offset) common_max_ptr = max(inp_max_ptr, out_max_ptr) base_size = (common_max_ptr - common_min_ptr) + 1 storage = xp.zeros(base_size, dtype=gy.dtype) flatten_full_indices = xp.arange(base_size, dtype=index_dtype[itemsize]) out_maybe_overlap = _maybe_overlapping_memory(out_shape, out_strides) if out_maybe_overlap: out_indices = _stride_array(flatten_full_indices, out_shape, out_strides, out_storage_offset - common_min_ptr) _index_add(storage, out_indices, gy) else: storage_view = _stride_array(storage, out_shape, out_strides, out_storage_offset - common_min_ptr) storage_view[:] = gy[:] inp_maybe_overlap = _maybe_overlapping_memory(inp_shape, inp_strides) if inp_maybe_overlap: count = xp.zeros_like(storage) inp_indices = _stride_array(flatten_full_indices, inp_shape, inp_strides, inp_storage_offset - common_min_ptr) _index_add(count, inp_indices, xp.ones(1)) with np.errstate(divide='ignore', invalid='ignore'): storage /= count return _stride_array(storage, inp_shape, inp_strides, inp_storage_offset - common_min_ptr), def backward(self, target_input_indexes, grad_outputs): raise NotImplementedError def as_strided(x, shape, strides, storage_offset=None): """Create a new view of array with the given shape, strides, and offset. Args: x (tuple of :class:`~chainer.Variable` or :class:`numpy.ndarray` or \ :class:`cupy.ndarray`): The array pointing a memory buffer. Its view is totally ignored. shape (tuple of int): The shape of output. strides (tuple of int): The strides of output, given in the unit of steps. storage_offset (int): The offset between the head of allocated memory and the pointer of first element, given in the unit of steps. Returns: ~chainer.Variable: The strided variable. .. warning:: Users should be aware that this function potentially causes unintended side effects. See `numpy.lib.stride_tricks.as_strided`_ for the detail. .. note:: The backward algorithm is borrowed from `torch.Tensor.as_strided`. Therefore, the returned gradient of ``backward`` is *layout-agnostic* when ``x`` contains memory overlap. See notes in pytorch's source code (as_strided Backward and layout-aware/agnostic autograd) too. .. note:: In this function ``strides`` and ``storage_offset`` are given in the unit of steps instead of bytes. This specification differs from :func:`numpy.lib.stride_tricks.as_strided`. .. admonition:: Example >>> from chainer import functions as F, Variable >>> x = Variable(np.arange(4, dtype=np.float32)) >>> x variable([0., 1., 2., 3.]) >>> y = F.as_strided(x, (3, 2), (1, 1), 0) >>> y variable([[0., 1.], [1., 2.], [2., 3.]]) >>> y.grad = np.ones((3, 2), dtype=np.float32) >>> y.backward() >>> x.grad array([1., 2., 2., 1.], dtype=float32) .. _numpy.lib.stride_tricks.as_strided: https://docs.scipy.org/doc/numpy/reference/generated/\ numpy.lib.stride_tricks.as_strided.html """ return AsStrided(shape, strides, storage_offset).apply((x,))[0]
14,427
35.994872
79
py
chainer
chainer-master/chainer/functions/array/diagonal.py
import numpy from chainer import backend from chainer import function_node from chainer.utils import type_check class Diagonal(function_node.FunctionNode): def __init__(self, offset, axis1, axis2): self.offset = offset self.axis1 = axis1 self.axis2 = axis2 def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) in_type = in_types[0] type_check.expect(max(self.axis1, self.axis2) < in_type.ndim) type_check.expect(-in_type.ndim <= min(self.axis1, self.axis2)) def forward(self, inputs): x, = inputs self._in_shape = x.shape y = x.diagonal(offset=self.offset, axis1=self.axis1, axis2=self.axis2) return y, def backward(self, indexes, grad_outputs): return DiagonalGrad( self._in_shape, self.offset, self.axis1, self.axis2 ).apply(grad_outputs) class DiagonalGrad(function_node.FunctionNode): def __init__(self, out_shape, offset, axis1, axis2): self.out_shape = out_shape self.offset = offset self.axis1 = axis1 self.axis2 = axis2 def forward(self, inputs): x, = inputs xp = backend.get_array_module(x) y = xp.zeros(self.out_shape, x.dtype) y_diag = y.diagonal( offset=self.offset, axis1=self.axis1, axis2=self.axis2) if xp is numpy: y_diag.setflags(write=True) y_diag[...] = x return y, def backward(self, indexes, grad_outputs): return Diagonal(self.offset, self.axis1, self.axis2).apply( grad_outputs) def diagonal(x, offset=0, axis1=0, axis2=1): """Take diagonal Axes other than ``axis1`` and ``axis2`` are regarded as batch dimensions. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): A variable to be sliced. offset (int): Offset from the principal diagonal. An upper diagonal matrix can have nonzero diagonals with nonnegative offsets. axis1 (int): First axis (that has row indices) of matrix axis2 (int): Second axis (that has column indices) of matrix Returns: ~chainer.Variable: (Batched) diagonal vectors .. admonition:: Example >>> x = chainer.Variable(np.arange(9).reshape(3, 3).astype(np.float32)) >>> x variable([[0., 1., 2.], [3., 4., 5.], [6., 7., 8.]]) >>> chainer.functions.diagonal(x, offset=1) variable([1., 5.]) """ return Diagonal(offset, axis1, axis2).apply((x,))[0]
2,584
29.411765
79
py
chainer
chainer-master/chainer/functions/array/fliplr.py
from chainer import backend from chainer import function_node from chainer.utils import type_check class FlipLR(function_node.FunctionNode): """Flip array in the left/right direction.""" def check_type_forward(self, in_types): type_check._argname(in_types, ('a',)) a_type = in_types[0] type_check.expect( a_type.dtype.kind == 'f', a_type.ndim >= 2 ) def forward(self, inputs): xp = backend.get_array_module(*inputs) return xp.fliplr(inputs[0]), def backward(self, indexes, grad_outputs): return FlipLR().apply(grad_outputs) def fliplr(a): """Flip array in the left/right direction. Args: a (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. Returns: ~chainer.Variable: Output variable. """ return FlipLR().apply((a,))[0]
878
22.131579
73
py
chainer
chainer-master/chainer/functions/array/moveaxis.py
import six from chainer import backend from chainer import function_node from chainer.utils import type_check def _normalize_axis_tuple(axis, ndim): ret = [] for ax in axis: ret.append(ax % ndim) return ret def _moveaxis(a, source, destination, xp): if hasattr(xp, 'moveaxis'): return xp.moveaxis(a, source, destination) if not all(isinstance(axis, six.integer_types) for axis in source): raise TypeError('int or tuple of int are required.') if not all(isinstance(axis, six.integer_types) for axis in destination): raise TypeError('int or tuple of int are required.') if len(source) != len(destination): raise ValueError('Length of source and destination are ' 'different.') source = _normalize_axis_tuple(source, a.ndim) destination = _normalize_axis_tuple(destination, a.ndim) if len(set(source)) != len(source): raise ValueError('duplicate value in source axis: ({})'.format( ', '.join(map(str, source)))) if len(set(destination)) != len(destination): raise ValueError('duplicate value in destination axis: ({})' .format(', '.join(map(str, destination)))) order = [n for n in six.moves.range(a.ndim) if n not in source] for dest, src in sorted(six.moves.zip(destination, source)): order.insert(dest, src) result = a.transpose(order) return result class Moveaxis(function_node.FunctionNode): """Move axis of an array.""" def __init__(self, source, destination): if isinstance(source, int): self.source = (source,) else: self.source = source if isinstance(destination, int): self.destination = (destination,) else: self.destination = destination def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) type_check.expect( in_types[0].dtype.kind == 'f', ) if self.source is not None: for axis in self.source: if axis >= 0: type_check.expect( axis < in_types[0].ndim, ) else: type_check.expect( -axis - 1 < in_types[0].ndim, ) if self.destination is not None: for axis in self.destination: if axis >= 0: type_check.expect( axis < in_types[0].ndim, ) else: type_check.expect( -axis - 1 < in_types[0].ndim, ) def forward(self, inputs): self.retain_inputs(()) self._in_ndim = inputs[0].ndim xp = backend.get_array_module(*inputs) return _moveaxis(inputs[0], self.source, self.destination, xp), def backward(self, indexes, gy): return Moveaxis(self.destination, self.source).apply(gy) def moveaxis(x, source, destination): """Move the source axes to the destination. This function transpose the input ``x`` by moving the axes ``source`` to the axes ``destination``. Other axes remain in their original order. See also :func:`chainer.functions.transpose`, :func:`chainer.functions.swapaxes`. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. source (int or tuple of int): Original positions of the axes to move. These must be unique. destination (int or tuple of int): Destination positions for each of the original axes. These must also be unique. Returns: ~chainer.Variable: Variable whose axis is moved. .. admonition:: Example >>> x = np.zeros((2, 3, 4, 5), np.float32) >>> chainer.functions.moveaxis(x, 0, -1).shape (3, 4, 5, 2) >>> chainer.functions.moveaxis(x, (0, 3), (2, 0)).shape (5, 3, 2, 4) """ return Moveaxis(source, destination).apply((x,))[0]
4,098
31.023438
76
py
chainer
chainer-master/chainer/functions/array/scatter_add.py
import numpy import chainer from chainer import backend from chainer.backends import cuda from chainer import function_node from chainer.utils import type_check import chainerx class ScatterAdd(function_node.FunctionNode): def __init__(self, slices): if isinstance(slices, list): if all([isinstance(s, int) for s in slices]): slices = slices, slices = tuple(slices) elif not isinstance(slices, tuple): slices = slices, if chainer.is_debug(): n_ellipses = 0 for s in slices: if s is Ellipsis: n_ellipses += 1 if n_ellipses > 1: raise ValueError('Only one Ellipsis is allowed') self.slices = slices def check_type_forward(self, in_types): type_check._argname(in_types, ('a', 'b')) n_nones = len([item for item in self.slices if item is None]) valid_slice = len(self.slices) - n_nones type_check.expect(in_types[0].ndim >= valid_slice) def forward(self, xs): a = xs[0] b = xs[1] y = a.copy() xp = backend.get_array_module(a) slices = tuple([ backend.from_chx(s) if isinstance(s, chainerx.ndarray) else s for s in self.slices]) if y[slices].shape != b.shape: raise ValueError( 'Chainer does not support automatic broadcasting ' 'of variables.') if xp is numpy: numpy.add.at(y, slices, b), else: cuda.cupyx.scatter_add(y, slices, b), return y, def backward(self, indexes, grad_outputs): gy, = grad_outputs ret = [] if 0 in indexes: ret.append(gy) if 1 in indexes: ret.append(gy[self.slices]) return ret def scatter_add(a, slices, b): """Adds given values to specified elements of an array. This function adds ``b`` to the specified elements of the copy of ``a``, and returns the copy. The value of the original ``a`` is not changed. Args: a (:class:`~chainer.Variable` or :ref:`ndarray`): A variable. slices (int, slice, Ellipsis, None, integer array-like, boolean\ array-like or tuple of them): It is an integer, a slice, an ellipsis, a numpy.newaxis, an integer array-like, a boolean array-like or tuple of them. b (:class:`~chainer.Variable` or :ref:`ndarray`): A variable that is scatter added to ``a``. Its shape has to equal ``a[slices]`` because broadcasting of variables is not supported. Returns: A :class:`~chainer.Variable` object which is the result of scatter addition. .. note:: It only supports types that are supported by CUDA's atomicAdd when an integer array is included in ``slices``. The supported types are ``numpy.float32``, ``numpy.int32``, ``numpy.uint32``, ``numpy.uint64`` and ``numpy.ulonglong``. .. note:: It does not support ``slices`` that contains multiple boolean arrays. .. seealso:: :func:`numpy.add.at` and :func:`cupyx.scatter_add`. """ y, = ScatterAdd(slices).apply((a, b)) return y
3,309
29.648148
77
py
chainer
chainer-master/chainer/functions/array/tile.py
import six import chainer from chainer import backend from chainer import function_node from chainer.utils import type_check class Tile(function_node.FunctionNode): """Tiling of an array.""" def __init__(self, reps): if isinstance(reps, six.integer_types): self.reps = (reps,) elif isinstance(reps, tuple) and all( isinstance(x, six.integer_types) for x in reps): self.reps = reps else: msg = 'reps must be int or tuple of ints.\n' \ 'Actual: {0}'.format(type(reps)) raise TypeError(msg) if not all(x >= 0 for x in self.reps): raise ValueError('All elements in reps must be zero or larger') def check_type_forward(self, in_types): type_check.expect(in_types.size() == 1) def forward(self, inputs): self._in_shape = inputs[0].shape xp = backend.get_array_module(*inputs) return xp.tile(inputs[0], self.reps), def backward(self, indexes, grad_outputs): reps = self.reps shape = tuple(self._in_shape) ndim = len(shape) # Ensure input and reps have the same length. if ndim > len(reps): reps = (1,) * (ndim - len(reps)) + reps elif ndim < len(reps): shape = (1,) * (len(reps) - ndim) + shape gy, = grad_outputs # Reshape so that base axis and reps axis can be distinguished. new_shape = [] for i in range(gy.ndim): new_shape.append(reps[i]) new_shape.append(shape[i]) new_shape = tuple(new_shape) # Sum along reps axis reps_axis = tuple(range(0, 2 * gy.ndim, 2)) gy = gy.reshape(new_shape) gy = chainer.functions.sum(gy, axis=reps_axis) if ndim < len(reps): return gy.reshape(self._in_shape), else: return gy, def tile(x, reps): """Construct an array by tiling a given array. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. Let the length of ``reps`` be ``d``. If ``x.ndim < d``, ``x`` is treated as ``d``-dimensional array by prepending new axes. For example, when the shape of ``x`` is ``(2,)`` and tiled with 2-dim repetitions, ``x`` is treated as the shape ``(1, 2)``. If ``x.ndim > d``, ``reps`` is treated as ``x.ndim``-dimensional by pre-pending 1's. For example, when the shape of ``x`` is ``(2, 3, 2, 3)``, the 2-dim ``reps`` of ``(2, 2)`` is treated as ``(1, 1, 2, 2)``. reps (:class:`int` or :class:`tuple` of :class:`int` s): The number of times which ``x`` is replicated along each axis. Returns: ~chainer.Variable: The tiled output Variable. Let the length of ``reps`` be ``d``, the output has the dimension of ``max(d, x.ndim)``. .. admonition:: Example >>> x = np.array([0, 1, 2]) >>> x.shape (3,) >>> y = F.tile(x, 2) >>> y.shape (6,) >>> y.array array([0, 1, 2, 0, 1, 2]) >>> y = F.tile(x, (2, 2)) >>> y.shape (2, 6) >>> y.array array([[0, 1, 2, 0, 1, 2], [0, 1, 2, 0, 1, 2]]) >>> y = F.tile(x, (2, 1, 2)) >>> y.shape (2, 1, 6) >>> y.array array([[[0, 1, 2, 0, 1, 2]], <BLANKLINE> [[0, 1, 2, 0, 1, 2]]]) >>> x = np.array([[1, 2], [3, 4]]) >>> x.shape (2, 2) >>> y = F.tile(x, 2) >>> y.shape (2, 4) >>> y.array array([[1, 2, 1, 2], [3, 4, 3, 4]]) >>> y = F.tile(x, (2, 2)) >>> y.shape (4, 4) >>> y.array array([[1, 2, 1, 2], [3, 4, 3, 4], [1, 2, 1, 2], [3, 4, 3, 4]]) >>> y = F.tile(x, (2, 1, 2)) >>> y.shape (2, 2, 4) >>> y.array array([[[1, 2, 1, 2], [3, 4, 3, 4]], <BLANKLINE> [[1, 2, 1, 2], [3, 4, 3, 4]]]) """ return Tile(reps).apply((x,))[0]
4,207
28.843972
78
py
chainer
chainer-master/chainer/functions/array/cast.py
import numpy import chainer from chainer import function_node from chainer.utils import type_check class Cast(function_node.FunctionNode): """Cast function.""" def __init__(self, typ): self.type = typ def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) def forward(self, x): self._in_type = x[0].dtype.type return x[0].astype(self.type, copy=False), def backward(self, indexes, g): if numpy.dtype(self._in_type).kind != 'f': gx = None else: gx = cast(g[0], self._in_type) return gx, def cast(x, typ): """Cast an input variable to a given type. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable to be casted. A \ :math:`(s_1, s_2, ..., s_N)`-shaped array. typ (:class:`str` of dtype or :class:`numpy.dtype`): Typecode or data type to cast. Returns: ~chainer.Variable: Variable holding a casted array. .. admonition:: Example >>> x = np.arange(0, 3, dtype=np.float64) >>> x.dtype dtype('float64') >>> y = F.cast(x, np.float32) >>> y.dtype dtype('float32') >>> y = F.cast(x, 'float16') >>> y.dtype dtype('float16') """ if x.dtype == typ: if not chainer.config.enable_backprop: return chainer.as_variable(x) return Cast(typ).apply((x,))[0]
1,478
23.65
60
py
chainer
chainer-master/chainer/functions/array/swapaxes.py
from chainer import function_node from chainer.utils import type_check class Swapaxes(function_node.FunctionNode): """Swap two axes of an array.""" def __init__(self, axis1, axis2): self.axis1 = axis1 self.axis2 = axis2 def check_type_forward(self, in_types): type_check.expect(in_types.size() == 1,) @property def label(self): return 'Swapaxes' def forward(self, inputs): x, = inputs return x.swapaxes(self.axis1, self.axis2), def backward(self, indexes, grad_outputs): gy, = grad_outputs return Swapaxes(self.axis1, self.axis2).apply((gy,)) def swapaxes(x, axis1, axis2): """Swap two axes of a variable. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`(s_1, s_2, ..., s_N)` -shaped float array. axis1 (int): The first axis to swap. axis2 (int): The second axis to swap. Returns: ~chainer.Variable: Variable whose axes are swapped. .. admonition:: Example >>> x = np.array([[[0, 1, 2], [3, 4, 5]]], np.float32) >>> x.shape (1, 2, 3) >>> y = F.swapaxes(x, axis1=0, axis2=1) >>> y.shape (2, 1, 3) >>> y.array array([[[0., 1., 2.]], <BLANKLINE> [[3., 4., 5.]]], dtype=float32) """ y, = Swapaxes(axis1, axis2).apply((x,)) return y
1,426
24.482143
73
py
chainer
chainer-master/chainer/functions/array/broadcast.py
import six import chainer from chainer import backend from chainer import function_node from chainer.utils import type_check import chainerx class Broadcast(function_node.FunctionNode): """Function that broadcasts given arrays.""" def check_type_forward(self, in_types): type_check.expect(in_types.size() > 0) shapes = [t.shape for t in in_types] type_check.expect_broadcast_shapes(*shapes) def forward(self, inputs): self._xp = backend.get_array_module(*inputs) self._in_shapes = [x.shape for x in inputs] self._in_dtypes = [x.dtype for x in inputs] return tuple(self._xp.broadcast_arrays(*inputs)) def backward(self, indexes, grad_outputs): return tuple([None if grad_outputs[i] is None else chainer.functions.sum_to( grad_outputs[i], self.inputs[i].shape) for i in indexes]) def broadcast(*args): """Broadcast given variables. Args: args (:class:`~chainer.Variable` or :ref:`ndarray`): Input variables to be broadcasted. Each dimension of the shapes \ of the input variables must have the same size. Returns: ~chainer.Variable: :class:`~chainer.Variable` or tuple of \ :class:`~chainer.Variable` objects which are broadcasted \ from the given arguments. .. admonition:: Example >>> x = np.random.uniform(0, 1, (3, 2)).astype(np.float32) >>> y = F.broadcast(x) >>> np.all(x == y.array) True >>> z = np.random.uniform(0, 1, (3, 2)).astype(np.float32) >>> y, w = F.broadcast(x, z) >>> np.all(x == y.array) & np.all(z == w.array) True """ if len(args) == 1: return chainer.as_variable(args[0]) return Broadcast().apply(args) class BroadcastTo(function_node.FunctionNode): """Function that broadcasts an array to a new shape.""" def __init__(self, shape): self._shape = tuple(shape) def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) ndim = type_check.make_variable(len(self._shape), 'len(shape)') type_check.expect(in_types[0].ndim <= ndim) shape = type_check.eval(in_types[0].shape) # check the shape in inverse order for i in six.moves.range(-1, -len(shape) - 1, -1): if shape[i] == self._shape[i] or shape[i] == 1: continue expect = 'in_type[0].shape[%d] == %d' % (i, self._shape[i]) if self._shape[i] != 1: expect += ' or in_type[0].shape[%d] == 1' % i actual = 'in_type[0].shape: %s' % str(shape) raise type_check.InvalidType(expect, actual) def broadcast_to(self, inputs): x, = inputs return chainerx.broadcast_to(x, self.shape), def forward(self, inputs): x, = inputs xp = backend.get_array_module(x) if hasattr(xp, 'broadcast_to'): return xp.broadcast_to(x, self._shape), else: # numpy 1.9 doesn't support broadcast_to method dummy = xp.empty(self._shape) bx, _ = xp.broadcast_arrays(x, dummy) return bx, def backward(self, indexes, grad_outputs): gx, = grad_outputs x_node, = self.inputs return chainer.functions.sum_to(gx, x_node.shape), def broadcast_to(x, shape): """Broadcast a given variable to a given shape. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable to be broadcasted. A \ :math:`(s_1, s_2, ..., s_N)`-shaped float array. shape (tuple): Tuple of :class:`int` of the shape of the \ output variable. Returns: ~chainer.Variable: Output variable broadcasted to the given shape. .. admonition:: Example >>> x = np.arange(0, 3) >>> x array([0, 1, 2]) >>> y = F.broadcast_to(x, (3, 3)) >>> y.array array([[0, 1, 2], [0, 1, 2], [0, 1, 2]]) """ if x.shape == shape: return chainer.as_variable(x) y, = BroadcastTo(shape).apply((x,)) return y
4,232
29.673913
77
py
chainer
chainer-master/chainer/functions/array/space2depth.py
import chainer from chainer import backend from chainer import function_node from chainer.utils import type_check class Space2Depth(function_node.FunctionNode): """Space to depth transformation.""" def __init__(self, r): self.r = r def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) type_check.expect( in_types[0].dtype.kind == 'f', in_types[0].ndim == 4 ) def forward(self, inputs): X, = inputs xp = backend.get_array_module(X) bsize, c, a, b = X.shape X = xp.reshape( X, (bsize, c, a // self.r, self.r, b // self.r, self.r)) X = xp.transpose(X, (0, 3, 5, 1, 2, 4)) X = xp.reshape( X, (bsize, self.r ** 2 * c, a // self.r, b // self.r)) return X, def backward(self, indexes, grad_outputs): gy, = grad_outputs gy = chainer.functions.depth2space(gy, self.r) return gy, def space2depth(X, r): """Computes the space2depth transformation for subpixel calculations. Args: X (:class:`~chainer.Variable` or :ref:`ndarray`): Variable holding a 4d array of shape ``(batch, channel, dim1 * r, dim2 * r)``. r (int): the downscaling factor. Returns: ~chainer.Variable: A variable holding the downscaled layer array from subpixel array sampling. The shape is ``(batch, channel * r * r, dim1, dim2)``. .. note:: This can be used to compute inverse super-resolution transformations. See https://arxiv.org/abs/1609.05158 for details. .. seealso:: :func:`depth2space` .. admonition:: Example >>> X = np.arange(24).reshape(1, 1, 4, 6).astype(np.float32) >>> X.shape (1, 1, 4, 6) >>> X array([[[[ 0., 1., 2., 3., 4., 5.], [ 6., 7., 8., 9., 10., 11.], [12., 13., 14., 15., 16., 17.], [18., 19., 20., 21., 22., 23.]]]], dtype=float32) >>> y = F.space2depth(X, 2) >>> y.shape (1, 4, 2, 3) >>> y.array array([[[[ 0., 2., 4.], [12., 14., 16.]], <BLANKLINE> [[ 1., 3., 5.], [13., 15., 17.]], <BLANKLINE> [[ 6., 8., 10.], [18., 20., 22.]], <BLANKLINE> [[ 7., 9., 11.], [19., 21., 23.]]]], dtype=float32) """ return Space2Depth(r).apply((X,))[0]
2,545
28.604651
77
py
chainer
chainer-master/chainer/functions/array/reshape.py
import chainer from chainer import function_node from chainer.utils import type_check def _count_unknown_dims(shape): cnt = 0 for dim in shape: cnt += dim < 0 return cnt class Reshape(function_node.FunctionNode): """Reshapes an input array without copy.""" def __init__(self, shape): self.shape = shape self._cnt = _count_unknown_dims(shape) assert self._cnt <= 1 def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type, = in_types if self._cnt == 0: type_check.expect( type_check.prod(x_type.shape) == type_check.prod(self.shape)) else: known_size = 1 for s in self.shape: if s > 0: known_size *= s size_var = type_check.make_variable( known_size, 'known_size(=%d)' % known_size) type_check.expect( type_check.prod(x_type.shape) % size_var == 0) def forward_chainerx(self, inputs): x, = inputs return x.reshape(self.shape), def forward(self, inputs): x, = inputs return x.reshape(self.shape), def backward(self, indexes, grad_outputs): gx, = grad_outputs return reshape(gx, self.inputs[0].shape), def reshape(x, shape): """Reshapes an input variable without copy. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. shape (:class:`tuple` of :class:`int` s): Expected shape of the output array. The number of elements which the array of ``shape`` contains must be equal to that of input array. One shape dimension can be -1. In this case, the value is inferred from the length of the array and remaining dimensions. Returns: ~chainer.Variable: Variable that holds a reshaped version of the input variable. .. seealso:: :func:`numpy.reshape`, :func:`cupy.reshape` .. admonition:: Example >>> x = np.array([[1, 2, 3, 4], [5, 6, 7, 8]]) >>> y = F.reshape(x, (8,)) >>> y.shape (8,) >>> y.array array([1, 2, 3, 4, 5, 6, 7, 8]) >>> y = F.reshape(x, (4, -1)) # the shape of output is inferred >>> y.shape (4, 2) >>> y.array array([[1, 2], [3, 4], [5, 6], [7, 8]]) >>> y = F.reshape(x, (4, 3)) \ # the shape of input and output are not consistent Traceback (most recent call last): ... chainer.utils.type_check.InvalidType: Invalid operation is performed in: Reshape (Forward) <BLANKLINE> Expect: prod(x.shape) == prod((4, 3)) Actual: 8 != 12 """ if x.shape == shape: return chainer.as_variable(x) y, = Reshape(shape).apply((x,)) return y
2,918
28.19
77
py
chainer
chainer-master/chainer/functions/array/where.py
import numpy import chainer from chainer import backend from chainer import function_node from chainer.utils import type_check class Where(function_node.FunctionNode): """Choose elements depending on condition.""" def __init__(self, condition): self.condition = condition def check_type_forward(self, in_types): type_check._argname(in_types, ('x', 'y')) x_type, y_type = in_types condition = type_check._make_variable_from_array( # allow scalar `condition` chainer.utils.force_array(self.condition), 'condition') type_check.expect( condition.dtype == numpy.bool_, x_type.dtype == y_type.dtype, ) type_check.expect_broadcast_shapes( condition.shape, x_type.shape, y_type.shape) def forward(self, inputs): # may broadcast xp = backend.get_array_module(*inputs) x, y = inputs condition = self.condition return xp.where(condition, x, y), def backward(self, indexes, grad_outputs): condition = self.condition xp = backend.get_array_module(condition) g, = grad_outputs zero = xp.zeros((), dtype=g.dtype) ret = [] if 0 in indexes: gx, = Where(condition).apply((g, zero)) ret.append(chainer.functions.sum_to(gx, self.inputs[0].shape)) if 1 in indexes: gy, = Where(condition).apply((zero, g)) ret.append(chainer.functions.sum_to(gy, self.inputs[1].shape)) return ret def where(condition, x, y): """Choose elements depending on condition. This function choose values depending on a given ``condition``. All ``condition``, ``x``, and ``y`` must have the same shape. Args: condition (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable containing the condition. A :math:`(s_1, s_2, ..., s_N)` -shaped boolean array. Only boolean array is permitted. x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable chosen when ``condition`` is ``True``. A :math:`(s_1, s_2, ..., s_N)` -shaped float array. y (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable chosen when ``condition`` is ``False``. A :math:`(s_1, s_2, ..., s_N)` -shaped float array. Returns: ~chainer.Variable: Variable containing chosen values. .. admonition:: Example >>> cond = np.array([[1, 0], [0, 1]], dtype=np.bool) >>> cond array([[ True, False], [False, True]]) >>> x = np.array([[1, 2], [3, 4]], np.float32) >>> y = np.zeros((2, 2), np.float32) >>> F.where(cond, x, y).array array([[1., 0.], [0., 4.]], dtype=float32) """ if isinstance(condition, chainer.Variable): condition = condition.array z, = Where(condition).apply((x, y)) return z
2,995
31.565217
74
py
chainer
chainer-master/chainer/functions/array/expand_dims.py
import chainer from chainer import backend from chainer import function_node from chainer.utils import type_check class ExpandDims(function_node.FunctionNode): """Expands dimensions of an input array without copy.""" def __init__(self, axis): self.axis = int(axis) def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type, = in_types if self.axis >= 0: type_check.expect(x_type.ndim >= self.axis) else: type_check.expect(x_type.ndim >= -self.axis - 1) def forward(self, inputs): x, = inputs xp = backend.get_array_module(x) return xp.expand_dims(x, self.axis), def backward(self, indexes, grad_outputs): gx, = grad_outputs return chainer.functions.reshape(gx, self.inputs[0].shape), def expand_dims(x, axis): """Expands dimensions of an input variable without copy. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. axis (int): Position where new axis is to be inserted. The ``axis`` parameter is acceptable when :math:`-ndim - 1 \\leq axis \\leq ndim`. (``ndim`` is the dimension of input variables). When :math:`axis < 0`, the result is the same with :math:`ndim + 1 - |axis|`. Returns: ~chainer.Variable: Variable that holds an expanded input. The ``ndim`` of output is one greater than that of ``x``. .. admonition:: Example >>> x = np.array([1, 2, 3]) >>> x.shape (3,) >>> y = F.expand_dims(x, axis=0) >>> y.shape (1, 3) >>> y.array array([[1, 2, 3]]) >>> y = F.expand_dims(x, axis=1) >>> y.shape (3, 1) >>> y.array array([[1], [2], [3]]) >>> y = F.expand_dims(x, axis=-2) >>> y.shape (1, 3) >>> y.array array([[1, 2, 3]]) """ y, = ExpandDims(axis).apply((x,)) return y
2,057
26.810811
78
py
chainer
chainer-master/chainer/functions/array/flatten.py
import chainer def flatten(x): """Flatten a given array into one dimension. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. Returns: ~chainer.Variable: Output variable flatten to one dimension. .. note:: When you input a scalar array (i.e. the shape is ``()``), you can also get the one dimension array whose shape is ``(1,)``. .. admonition:: Example >>> x = np.array([[1, 2], [3, 4]]) >>> x.shape (2, 2) >>> y = F.flatten(x) >>> y.shape (4,) >>> y.array array([1, 2, 3, 4]) >>> x = np.arange(8).reshape(2, 2, 2) >>> x.shape (2, 2, 2) >>> y = F.flatten(x) >>> y.shape (8,) >>> y.array array([0, 1, 2, 3, 4, 5, 6, 7]) """ return chainer.functions.reshape(x, (x.size,))
891
21.3
73
py
chainer
chainer-master/chainer/functions/array/hstack.py
import numpy import six import chainer from chainer import backend from chainer import function_node from chainer.utils import type_check class Hstack(function_node.FunctionNode): """Concatenate multiple tensors horizontally (column wise).""" def check_type_forward(self, in_types): type_check.expect(in_types.size() > 0) type_check._argname((in_types[0],), ('x0',)) ndim = type_check.eval(in_types[0].ndim) for i in six.moves.range(1, type_check.eval(in_types.size())): type_check._argname((in_types[i],), ('x{}'.format(i),)) type_check.expect( in_types[0].dtype == in_types[i].dtype, in_types[0].ndim == in_types[i].ndim, ) if ndim <= 1: continue for d in six.moves.range(0, ndim): if d == 1: continue type_check.expect(in_types[0].shape[d] == in_types[i].shape[d]) def forward(self, xs): xp = backend.get_array_module(*xs) return xp.hstack(xs), def backward(self, indexes, grad_outputs): gy, = grad_outputs ndim = len(self.inputs[0].shape) if len(self.inputs) == 1: if ndim == 0: return gy.reshape(()), return gy, if ndim == 0: gx = chainer.functions.split_axis(gy, len(self.inputs), 0) return [g.reshape(()) for g in gx] axis = 0 if ndim == 1 else 1 sizes = numpy.array([x.shape[axis] for x in self.inputs[:-1]]).cumsum() return chainer.functions.split_axis(gy, sizes, axis) def hstack(xs): """Concatenate variables horizontally (column wise). Args: xs (list of :class:`~chainer.Variable` or :ref:`ndarray`): Input variables to be concatenated. The variables must have the same ``ndim``. When the variables have the second axis (i.e. :math:`ndim \\geq 2`), the variables must have the same shape along all but the second axis. When the variables do not have the second axis(i.e. :math:`ndim < 2`), the variables need not to have the same shape. Returns: ~chainer.Variable: Output variable. When the input variables have the second axis (i.e. :math:`ndim \\geq 2`), the shapes of inputs and output are the same along all but the second axis. The length of second axis is the sum of the lengths of inputs' second axis. When the variables do not have the second axis (i.e. :math:`ndim < 2`), the shape of output is ``(N, )`` (``N`` is the sum of the input variables' size). .. admonition:: Example >>> x1 = np.array((1, 2, 3)) >>> x1.shape (3,) >>> x2 = np.array((2, 3, 4)) >>> x2.shape (3,) >>> y = F.hstack((x1, x2)) >>> y.shape (6,) >>> y.array array([1, 2, 3, 2, 3, 4]) >>> x1 = np.arange(0, 12).reshape(3, 4) >>> x1.shape (3, 4) >>> x1 array([[ 0, 1, 2, 3], [ 4, 5, 6, 7], [ 8, 9, 10, 11]]) >>> x2 = np.arange(12, 18).reshape(3, 2) >>> x2.shape (3, 2) >>> x2 array([[12, 13], [14, 15], [16, 17]]) >>> y = F.hstack([x1, x2]) >>> y.shape (3, 6) >>> y.array array([[ 0, 1, 2, 3, 12, 13], [ 4, 5, 6, 7, 14, 15], [ 8, 9, 10, 11, 16, 17]]) """ y, = Hstack().apply(xs) return y
3,651
31.035088
79
py
chainer
chainer-master/chainer/functions/array/transpose.py
import numpy from chainer import function_node from chainer.utils import type_check class Transpose(function_node.FunctionNode): """Permute the dimensions of an array.""" def __init__(self, axes=None): self.axes = axes def check_type_forward(self, in_types): type_check.expect(in_types.size() == 1,) @property def label(self): return 'Transpose' def forward_chainerx(self, inputs): x = inputs[0] return x.transpose(self.axes), def forward(self, inputs): x = inputs[0] return x.transpose(self.axes), def backward(self, indexes, grad_outputs): inv_axes = self.axes if inv_axes: axes_len = len(inv_axes) inv_axes = tuple(numpy.argsort([ax % axes_len for ax in inv_axes])) return Transpose(inv_axes).apply(grad_outputs) def transpose(x, axes=None): """Permute the dimensions of an input variable without copy. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable to be transposed. A :math:`(s_1, s_2, ..., s_N)` -shaped float array. axes (tuple of ints): By default, reverse the dimensions, otherwise permute the axes according to the values given. Returns: ~chainer.Variable: Variable whose axes are permuted. .. admonition:: Example >>> x = np.array([[[0, 1, 2], [3, 4, 5]]], np.float32) >>> x.shape (1, 2, 3) >>> y = F.transpose(x) # reverse the dimensions >>> y.shape (3, 2, 1) >>> y.array array([[[0.], [3.]], <BLANKLINE> [[1.], [4.]], <BLANKLINE> [[2.], [5.]]], dtype=float32) >>> y = F.transpose(x, axes=(1, 0, 2)) # swap 1st and 2nd axis >>> y.shape (2, 1, 3) >>> y.array array([[[0., 1., 2.]], <BLANKLINE> [[3., 4., 5.]]], dtype=float32) """ return Transpose(axes).apply((x,))[0]
2,061
26.131579
79
py
chainer
chainer-master/chainer/functions/array/spatial_transformer_grid.py
import numpy import chainer from chainer import backend from chainer.backends import cuda from chainer import function from chainer.utils import argument from chainer.utils import type_check if cuda.cudnn_enabled: cudnn = cuda.cudnn libcudnn = cuda.libcudnn _sampler_type = cuda.libcudnn.CUDNN_SAMPLER_BILINEAR class SpatialTransformerGrid(function.Function): def __init__(self, output_shape): self.output_shape = output_shape def check_type_forward(self, in_types): type_check._argname(in_types, ('theta',)) theta_type = in_types[0] type_check.expect( theta_type.dtype.kind == 'f', theta_type.ndim == 3, theta_type.shape[1] == 2, theta_type.shape[2] == 3, ) def forward_cpu(self, inputs): return self._forward(inputs) def forward_gpu(self, inputs): if not chainer.should_use_cudnn('>=auto', 5000): return self._forward(inputs) theta, = inputs B, _, _ = theta.shape H, W = self.output_shape grid_t = cuda.cupy.empty((B, H, W, 2), dtype=theta.dtype) # Unlike spatial_transformer_sampler, # channel size can be anything in this case. shape = numpy.array((B, 1, H, W), dtype=numpy.int32) theta = cuda.cupy.ascontiguousarray(theta) handle = cudnn.get_handle() self.st_desc =\ cuda.cupy.cudnn.create_spatial_transformer_descriptor( _sampler_type, grid_t.dtype, len(shape), shape.ctypes.data) libcudnn.spatialTfGridGeneratorForward( handle, self.st_desc.value, theta.data.ptr, grid_t.data.ptr) grid = cuda.cupy.transpose(grid_t, (0, 3, 1, 2)) return grid, def _forward(self, inputs): theta, = inputs H, W = self.output_shape B, _, _ = theta.shape xp = backend.get_array_module(theta) ys, xs = xp.meshgrid( xp.linspace(-1, 1, H, dtype=theta.dtype), xp.linspace(-1, 1, W, dtype=theta.dtype), indexing='ij', copy=False ) coords = xp.concatenate( [xs[None], ys[None], xp.ones((1, H, W), dtype=theta.dtype)], axis=0) grid = theta.dot(coords.reshape(3, H * W)).reshape(B, 2, H, W) return grid, def backward_cpu(self, inputs, grad_outputs): return self._backward(inputs, grad_outputs) def backward_gpu(self, inputs, grad_outputs): if not chainer.should_use_cudnn('>=auto', 5000): return self._backward(inputs, grad_outputs) theta, = inputs ggrid, = grad_outputs ggrid_t = cuda.cupy.transpose(ggrid, (0, 2, 3, 1)) gtheta = cuda.cupy.empty_like(theta) handle = cudnn.get_handle() ggrid_t = cuda.cupy.ascontiguousarray(ggrid_t) libcudnn.spatialTfGridGeneratorBackward( handle, self.st_desc.value, ggrid_t.data.ptr, gtheta.data.ptr) return gtheta, def _backward(self, inputs, grad_outputs): theta, = inputs ggrid, = grad_outputs H, W = self.output_shape B, _, _ = theta.shape xp = backend.get_array_module(theta) ys, xs = xp.meshgrid( xp.linspace(-1, 1, H, dtype=theta.dtype), xp.linspace(-1, 1, W, dtype=theta.dtype), indexing='ij', copy=False ) coords = xp.concatenate( [xs[None], ys[None], xp.ones((1, H, W), dtype=theta.dtype)], axis=0) coords_T = coords.reshape(3, H * W).transpose(1, 0) ggrid = ggrid.reshape(B, 2, H * W) gtheta = ggrid.dot(coords_T).reshape(B, 2, 3) return gtheta, def spatial_transformer_grid(theta, output_shape, **kwargs): """2D Spatial Transformer grid. This function generates coordinates of the points sampled from an image to perform warping described in `Spatial Transformer Networks <https://arxiv.org/abs/1506.02025>`_. Given a coordinate in the warped image :math:`(x_i^t, y_i^t)`, the point sampled from the source image :math:`(x_i^s, y_i^s)` are calculated by the following equation. .. note:: cuDNN supports SpatialTransformerGrid from version 5.0.0. .. math:: \\left(\\begin{matrix} x_i^s \\\\ y_i^s \\end{matrix}\\right) = \\left(\\begin{matrix} \\theta_{11} & \\theta_{12} & \\theta_{13} \\\\ \\theta_{21} & \\theta_{22} & \\theta_{23} \\end{matrix}\\right) \\left(\\begin{matrix} x_i^t \\\\ y_i^t \\\\ 1 \\end{matrix}\\right) Notation: here is a notation for dimensionalities. - :math:`n` is the batch size. - :math:`h_O` and :math:`w_O` are the height and the width of the output image. Args: theta (:class:`~chainer.Variable` or :ref:`ndarray`): An array of shape :math:`(n, 2, 3)`. This is a batch of :math:`2 \\times 3` matrix used for the warping described above. output_shape (tuple): A tuple of 2 elements: :math:`h_O, w_O`. Returns: ~chainer.Variable: A variable of shape :math:`(n, 2, h_O, w_O)`. In the 2nd dimension, the first element is the coordinate along the x axis, and the second element is the coordinate along the y axis. All the coordinates in the image are scaled to fit range :math:`[-1, 1]`. This means that the coordinate :math:`(-1, -1)` corresponds to the upper-left corner of the input image. """ if kwargs: argument.check_unexpected_kwargs( kwargs, use_cudnn='The argument "use_cudnn" is not ' 'supported anymore. ' 'Use chainer.using_config(\'use_cudnn\', value) ' 'context where value can be `always`, `never`, or `auto`.') argument.assert_kwargs_empty(kwargs) return SpatialTransformerGrid(output_shape)(theta)
5,933
33.5
78
py
chainer
chainer-master/chainer/functions/array/repeat.py
import six from chainer import backend from chainer import function_node from chainer import utils from chainer.utils import type_check class Repeat(function_node.FunctionNode): """Repeat elements of an array.""" def __init__(self, repeats, axis=None): if isinstance(repeats, six.integer_types): self.repeats = (repeats,) elif isinstance(repeats, tuple) and all( isinstance(x, six.integer_types) for x in repeats): # Although it is not explicitly documented, NumPy/CuPy allows # specifying bool or tuple of bools as `repeats`. # Thus we just check type against `six.integer_types`, without # excluding `bool`. self.repeats = repeats else: raise TypeError('repeats must be int or tuple of ints') if not all(x >= 0 for x in self.repeats): raise ValueError('all elements in repeats must be zero or larger') if axis is not None and ( not isinstance(axis, six.integer_types) or isinstance(axis, bool)): # `axis` cannot be bool, in contrast to `repeats`. raise TypeError('axis must be int or None') self.axis = axis def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) def forward(self, inputs): self.retain_inputs((0,)) x, = inputs xp = backend.get_array_module(x) repeats = self.repeats # Workaround for bug in NumPy 1.9 that specifying one element list to # `repeats` fails to broadcast. if len(repeats) == 1: repeats = repeats[0] return xp.repeat(x, repeats, self.axis), def backward(self, indexes, grad_outputs): x, = self.get_retained_inputs() return RepeatGrad(self.repeats, self.axis, x.shape, x.dtype).apply( grad_outputs) class RepeatGrad(function_node.FunctionNode): def __init__(self, repeats, axis, in_shape, in_dtype): self.repeats = repeats self.axis = axis if axis is not None and axis < 0: self.axis += len(in_shape) self.in_shape = in_shape self.in_dtype = in_dtype def forward(self, inputs): gy, = inputs xp = backend.get_array_module(gy) repeats = self.repeats axis = self.axis shape = list(self.in_shape) dtype = self.in_dtype if len(gy) == 0: gx = xp.zeros(shape, dtype) return gx, if len(repeats) == 1: repeats = int(repeats[0]) if axis is None: gx = gy.reshape(-1, repeats).sum(axis=1).reshape(shape) else: shape[axis:axis + 1] = [-1, repeats] gx = gy.reshape(shape).sum(axis=axis + 1) return gx, if axis is None: pos = 0 gx = xp.zeros(utils.size_of_shape(shape), dtype) for (i, r) in enumerate(repeats): gx[i] = xp.sum(gy[pos:pos + r]) pos += r gx = gx.reshape(shape) else: gx = xp.zeros(shape, dtype) pos = 0 src = [slice(None)] * axis + [None] dst = [slice(None)] * axis + [None] for (i, r) in enumerate(repeats): src[-1] = slice(pos, pos + r) dst[-1] = slice(i, i + 1) gx[tuple(dst)] = gy[tuple(src)].sum(axis=axis, keepdims=True) pos += r return gx, def backward(self, indexes, grad_outputs): return Repeat(self.repeats, self.axis).apply(grad_outputs) def repeat(x, repeats, axis=None): """Construct an array by repeating a given array. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. repeats (:class:`int` or :class:`tuple` of :class:`int` s): The number of times which each element of ``x`` is repeated. axis (:class:`int`): The axis along which to repeat values. Returns: ~chainer.Variable: The repeated output Variable. .. admonition:: Example >>> x = np.array([0, 1, 2]) >>> x.shape (3,) >>> y = F.repeat(x, 2) >>> y.shape (6,) >>> y.array array([0, 0, 1, 1, 2, 2]) >>> x = np.array([[1,2], [3,4]]) >>> x.shape (2, 2) >>> y = F.repeat(x, 3, axis=1) >>> y.shape (2, 6) >>> y.array array([[1, 1, 1, 2, 2, 2], [3, 3, 3, 4, 4, 4]]) >>> y = F.repeat(x, (1, 2), axis=0) >>> y.shape (3, 2) >>> y.array array([[1, 2], [3, 4], [3, 4]]) """ return Repeat(repeats, axis).apply((x,))[0]
4,817
29.884615
78
py
chainer
chainer-master/chainer/functions/array/select_item.py
import numpy import six import chainer from chainer import backend from chainer.backends import cuda from chainer import function_node from chainer.utils import type_check class SelectItem(function_node.FunctionNode): """Select elements stored in given indices.""" def check_type_forward(self, in_types): type_check._argname(in_types, ('x', 't')) x_type, t_type = in_types type_check.expect( t_type.dtype.kind == 'i', x_type.ndim == 2, t_type.ndim == 1, x_type.shape[0] == t_type.shape[0], ) def forward(self, inputs): self.retain_inputs((1,)) x, t = inputs self._in_shape = x.shape self._in_dtype = x.dtype if chainer.is_debug(): if not ((0 <= t).all() and (t < x.shape[1]).all()): msg = 'Each label `t` need to satisfty `0 <= t < x.shape[1]`' raise ValueError(msg) xp = backend.get_array_module(x) if xp is numpy: # This code is equivalent to `t.choose(x.T)`, but `numpy.choose` # does not work when `x.shape[1] > 32`. return x[six.moves.range(t.size), t], else: y = cuda.elementwise( 'S t, raw T x', 'T y', 'int ind[] = {i, t}; y = x[ind];', 'getitem_fwd' )(t, x) return y, def backward(self, indexes, gy): t = self.get_retained_inputs()[0] ret = [] if 0 in indexes: gx = Assign(self._in_shape, self._in_dtype, t).apply(gy)[0] ret.append(gx) if 1 in indexes: ret.append(None) return ret class Assign(function_node.FunctionNode): def __init__(self, shape, dtype, t): self.shape = shape self.dtype = dtype self.t = t.data def forward_cpu(self, inputs): t = backend.from_chx(self.t) # Workaround for ChainerX. gx = numpy.zeros(self.shape, self.dtype) gx[six.moves.range(self.t.size), t] = inputs[0] return gx, def forward_gpu(self, inputs): t = backend.from_chx(self.t) # Workaround for ChainerX. gx = cuda.cupy.zeros(self.shape, self.dtype) gx = cuda.elementwise( 'S t, T gloss', 'raw T gx', 'int ind[] = {i, t}; gx[ind] = gloss;', 'getitem_bwd' )(t, inputs[0], gx) return gx, def backward(self, indexes, gy): return SelectItem().apply((gy[0], self.t)) def select_item(x, t): """Select elements stored in given indices. This function returns ``t.choose(x.T)``, that means ``y[i] == x[i, t[i]]`` for all ``i``. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Variable storing arrays. A two-dimensional float array. t (:class:`~chainer.Variable` or :ref:`ndarray`): Variable storing index numbers. A one-dimensional int array. Length of the ``t`` should be equal to ``x.shape[0]``. Returns: ~chainer.Variable: Variable that holds ``t``-th element of ``x``. .. admonition:: Example >>> x = np.array([[0, 1, 2], [3, 4, 5]], np.float32) >>> t = np.array([0, 2], np.int32) >>> y = F.select_item(x, t) >>> y.shape (2,) >>> y.array array([0., 5.], dtype=float32) """ return SelectItem().apply((x, t))[0]
3,484
28.041667
77
py
chainer
chainer-master/chainer/functions/array/im2col.py
import numpy from chainer import function_node from chainer.utils.conv import col2im_cpu from chainer.utils.conv import col2im_gpu from chainer.utils.conv import im2col_cpu from chainer.utils.conv import im2col_gpu from chainer.utils import type_check def _pair(x): if hasattr(x, '__getitem__'): return x return x, x def _col2im(x, *args, **kwargs): if isinstance(x, numpy.ndarray): return col2im_cpu(x, *args, **kwargs) return col2im_gpu(x, *args, **kwargs) def _im2col(x, *args, **kwargs): if isinstance(x, numpy.ndarray): return im2col_cpu(x, *args, **kwargs) return im2col_gpu(x, *args, **kwargs) class Im2Col(function_node.FunctionNode): """Im2Col function.""" def __init__(self, ksize, stride, pad, cover_all, dilate): self.kh, self.kw = _pair(ksize) self.sy, self.sx = _pair(stride) self.ph, self.pw = _pair(pad) self.dy, self.dx = _pair(dilate) self.cover_all = cover_all def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type = in_types[0] type_check.expect( x_type.dtype.kind == 'f', x_type.ndim == 4 ) def forward(self, inputs): x, = inputs y = _im2col( x, self.kh, self.kw, self.sy, self.sx, self.ph, self.pw, cover_all=self.cover_all, dy=self.dy, dx=self.dx) n, c, kh, kw, out_h, out_w = y.shape return y.reshape(n, c * kh * kw, out_h, out_w), def backward(self, indexes, grad_outputs): return Im2ColGrad((self.kh, self.kw), (self.sy, self.sx), (self.ph, self.pw), self.cover_all, (self.dy, self.dx), self.inputs[0].shape) \ .apply(grad_outputs) class Im2ColGrad(function_node.FunctionNode): """Im2Col gradient function.""" def __init__(self, ksize, stride, pad, cover_all, dilate, in_shape): self.kh, self.kw = _pair(ksize) self.sy, self.sx = _pair(stride) self.ph, self.pw = _pair(pad) self.dy, self.dx = _pair(dilate) self.cover_all = cover_all self.in_shape = in_shape def check_type_forward(self, in_types): type_check._argname(in_types, ('gy',)) gy_type = in_types[0] type_check.expect( gy_type.dtype.kind == 'f', gy_type.ndim == 4 ) def forward(self, inputs): _, c, h, w = self.in_shape gy, = inputs n, _, out_h, out_w = gy.shape gy = gy.reshape(n, c, self.kh, self.kw, out_h, out_w) gx = _col2im( gy, self.sy, self.sx, self.ph, self.pw, h, w, self.dy, self.dx) return gx, def backward(self, indexes, grad_outputs): return Im2Col( (self.kh, self.kw), (self.sy, self.sx), (self.ph, self.pw), self.cover_all, (self.dy, self.dx)).apply(grad_outputs) def im2col(x, ksize, stride=1, pad=0, cover_all=False, dilate=1): """Extract patches from an image based on the filter. This function rearranges patches of an image and puts them in the channel dimension of the output. Patches are extracted at positions shifted by multiples of ``stride`` from the first position ``-pad`` for each spatial axis. The right-most (or bottom-most) patches do not run over the padded spatial size. Notation: here is a notation. - :math:`n` is the batch size. - :math:`c` is the number of the input channels. - :math:`h` and :math:`w` are the height and width of the input image, respectively. - :math:`k_H` and :math:`k_W` are the height and width of the filters, respectively. - :math:`s_Y` and :math:`s_X` are the strides of the filter. - :math:`p_H` and :math:`p_W` are the spatial padding sizes. - :math:`d_Y` and :math:`d_X` are the dilation factors of filter \ application. The output size :math:`(h_O, w_O)` is determined by the following equations when ``cover_all = False``: .. math:: h_O &= (h + 2p_H - k_H - (k_H - 1) * (d_Y - 1)) / s_Y + 1,\\\\ w_O &= (w + 2p_W - k_W - (k_W - 1) * (d_X - 1)) / s_X + 1. When ``cover_all = True``, the output size is determined by the following equations: .. math:: h_O &= (h + 2p_H - k_H - (k_H - 1) * (d_Y - 1) + s_Y - 1) / s_Y + 1,\\\\ w_O &= (w + 2p_W - k_W - (k_W - 1) * (d_X - 1) + s_X - 1) / s_X + 1. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable of shape :math:`(n, c, h, w)`. ksize (int or pair of ints): Size of filters (a.k.a. kernels). ``ksize=k`` and ``ksize=(k, k)`` are equivalent. stride (int or pair of ints): Stride of filter applications. ``stride=s`` and ``stride=(s, s)`` are equivalent. pad (int or pair of ints): Spatial padding width for input arrays. ``pad=p`` and ``pad=(p, p)`` are equivalent. cover_all (bool): If ``True``, all spatial locations are rearranged into some output pixels. It may make the output size larger. dilate (int or pair of ints): Dilation factor of filter applications. ``dilate=d`` and ``dilate=(d, d)`` are equivalent. Returns: ~chainer.Variable: Output variable whose shape is :math:`(n, c \\cdot k_H \\cdot k_W, h_O, w_O)` """ return Im2Col(ksize, stride, pad, cover_all, dilate).apply((x,))[0]
5,495
32.717791
79
py
chainer
chainer-master/chainer/functions/array/vstack.py
import numpy import six import chainer from chainer import backend from chainer import function_node from chainer.utils import type_check class Vstack(function_node.FunctionNode): """Concatenate multiple tensors vertically (row wise).""" def check_type_forward(self, in_types): type_check.expect(in_types.size() > 0) ndim = type_check.eval(in_types[0].ndim) for i in six.moves.range(1, type_check.eval(in_types.size())): type_check.expect( in_types[0].dtype == in_types[i].dtype, in_types[0].ndim == in_types[i].ndim, ) if ndim <= 1: type_check.expect(in_types[0].shape == in_types[i].shape) continue for d in six.moves.range(1, ndim): type_check.expect(in_types[0].shape[d] == in_types[i].shape[d]) def forward(self, xs): xp = backend.get_array_module(*xs) return xp.vstack(xs), def backward(self, indexes, grad_outputs): gy, = grad_outputs ndim = len(self.inputs[0].shape) if len(self.inputs) == 1: if ndim <= 1: return gy.reshape(self.inputs[0].shape), return gy, if ndim <= 1: gxs = chainer.functions.split_axis(gy, len(self.inputs), 0) return [gx.reshape(self.inputs[0].shape) for gx in gxs] sizes = numpy.array([x.shape[0] for x in self.inputs[:-1]]).cumsum() return chainer.functions.split_axis(gy, sizes, 0) def vstack(xs): """Concatenate variables vertically (row wise). Args: xs (list of :class:`~chainer.Variable` or :ref:`ndarray`): Input variables to be concatenated. The variables must have the same ``ndim``. When the variables have the second axis (i.e. :math:`ndim \\geq 2`), the variables must have the same shape along all but the first axis. When the variables do not have the second axis(i.e. :math:`ndim < 2`), the variables must have the same shape. Returns: ~chainer.Variable: Output variable. When the input variables have the second axis (i.e. :math:`ndim \\geq 2`), the shapes of inputs and output are the same along all but the first axis. The length of first axis is the sum of the lengths of inputs' first axis. When the variables do not have the second axis (i.e. :math:`ndim < 2`), the shape of output is ``(2, N)`` (``N`` is the size of the input variable). .. admonition:: Example >>> x1 = np.array((1, 2, 3)) >>> x1.shape (3,) >>> x2 = np.array((2, 3, 4)) >>> x2.shape (3,) >>> y = F.vstack((x1, x2)) >>> y.shape (2, 3) >>> y.array array([[1, 2, 3], [2, 3, 4]]) >>> x1 = np.arange(0, 12).reshape(3, 4) >>> x1.shape (3, 4) >>> x1 array([[ 0, 1, 2, 3], [ 4, 5, 6, 7], [ 8, 9, 10, 11]]) >>> x2 = np.arange(12, 20).reshape(2, 4) >>> x2.shape (2, 4) >>> x2 array([[12, 13, 14, 15], [16, 17, 18, 19]]) >>> y = F.vstack([x1, x2]) >>> y.shape (5, 4) >>> y.array array([[ 0, 1, 2, 3], [ 4, 5, 6, 7], [ 8, 9, 10, 11], [12, 13, 14, 15], [16, 17, 18, 19]]) """ return Vstack().apply((xs))[0]
3,567
31.144144
79
py
chainer
chainer-master/chainer/functions/array/__init__.py
0
0
0
py
chainer
chainer-master/chainer/functions/array/stack.py
import chainer from chainer import backend from chainer import function_node from chainer.utils import type_check import chainerx class Stack(function_node.FunctionNode): """Concatenate variables along a new axis.""" def __init__(self, axis): self.axis = axis def check_type_forward(self, in_types): type_check.expect(in_types.size() > 0) type_check.expect( -in_types[0].ndim - 1 <= self.axis, self.axis <= in_types[0].ndim ) dtype = in_types[0].dtype shape = in_types[0].shape for x_type in in_types[1:]: type_check.expect( x_type.dtype == dtype, x_type.shape == shape, ) def forward(self, inputs): xp = backend.get_array_module(*inputs) if hasattr(xp, 'stack'): return xp.stack(inputs, axis=self.axis), else: # Old numpy does not have numpy.stack. return xp.concatenate( [xp.expand_dims(x, self.axis) for x in inputs], self.axis), def forward_chainerx(self, xs): return chainerx.stack(xs, self.axis), def backward(self, inputs, grads): return chainer.functions.separate(grads[0], self.axis) def stack(xs, axis=0): """Concatenate variables along a new axis. Args: xs (list of :class:`~chainer.Variable` or :ref:`ndarray`): Input variables to be concatenated. The variables must have the same shape. axis (int): The axis along which the arrays will be stacked. The ``axis`` parameter is acceptable when :math:`-ndim - 1 \\leq axis \\leq ndim`. (``ndim`` is the dimension of input variables). When :math:`axis < 0`, the result is the same with :math:`ndim + 1 - |axis|`. Returns: ~chainer.Variable: Output variable. Let ``x_1, x_2, ..., x_n`` and ``y`` be the input variables and the output variable, ``y[:, ..., 0, ..., :]`` is ``x_1``, ``y[:, ..., 1, ..., :]`` is ``x_2`` and ``y[:, ..., n-1, ..., :]`` is ``x_n`` (The indexed axis indicates the ``axis``). .. admonition:: Example >>> x1 = np.arange(0, 12).reshape(3, 4) >>> x1.shape (3, 4) >>> x1 array([[ 0, 1, 2, 3], [ 4, 5, 6, 7], [ 8, 9, 10, 11]]) >>> x2 = np.arange(12, 24).reshape(3, 4) >>> x2.shape (3, 4) >>> x2 array([[12, 13, 14, 15], [16, 17, 18, 19], [20, 21, 22, 23]]) >>> y = F.stack([x1, x2], axis=0) >>> y.shape (2, 3, 4) >>> y.array array([[[ 0, 1, 2, 3], [ 4, 5, 6, 7], [ 8, 9, 10, 11]], <BLANKLINE> [[12, 13, 14, 15], [16, 17, 18, 19], [20, 21, 22, 23]]]) >>> y = F.stack([x1, x2], axis=1) >>> y.shape (3, 2, 4) >>> y.array array([[[ 0, 1, 2, 3], [12, 13, 14, 15]], <BLANKLINE> [[ 4, 5, 6, 7], [16, 17, 18, 19]], <BLANKLINE> [[ 8, 9, 10, 11], [20, 21, 22, 23]]]) >>> y = F.stack([x1, x2], axis=2) >>> y.shape (3, 4, 2) >>> y.array array([[[ 0, 12], [ 1, 13], [ 2, 14], [ 3, 15]], <BLANKLINE> [[ 4, 16], [ 5, 17], [ 6, 18], [ 7, 19]], <BLANKLINE> [[ 8, 20], [ 9, 21], [10, 22], [11, 23]]]) >>> y = F.stack([x1, x2], axis=-1) >>> y.shape (3, 4, 2) """ return Stack(axis).apply(xs)[0]
3,888
28.915385
78
py
chainer
chainer-master/chainer/functions/array/dstack.py
import numpy import six import chainer from chainer import backend from chainer import function_node from chainer.utils import type_check class Dstack(function_node.FunctionNode): """Concatenate multiple tensors along third axis (depth wise).""" def check_type_forward(self, in_types): type_check.expect(in_types.size() > 0) type_check._argname((in_types[0],), ('x0',)) ndim = type_check.eval(in_types[0].ndim) for i in six.moves.range(1, type_check.eval(in_types.size())): type_check._argname((in_types[i],), ('x{}'.format(i),)) type_check.expect( in_types[0].dtype == in_types[i].dtype, in_types[0].ndim == in_types[i].ndim, ) if ndim <= 2: type_check.expect(in_types[0].shape == in_types[i].shape) continue for d in six.moves.range(0, ndim): if d == 2: continue type_check.expect(in_types[0].shape[d] == in_types[i].shape[d]) def forward(self, xs): xp = backend.get_array_module(*xs) return xp.dstack(xs), def backward(self, indexes, grad_outputs): gy, = grad_outputs ndim = len(self.inputs[0].shape) if len(self.inputs) == 1: if ndim <= 2: return gy.reshape(self.inputs[0].shape), return gy, if ndim <= 2: gxs = chainer.functions.split_axis(gy, len(self.inputs), axis=2) return [gx.reshape(self.inputs[0].shape) for gx in gxs] sizes = numpy.array([x.shape[2] for x in self.inputs[:-1]]).cumsum() return chainer.functions.split_axis(gy, sizes, axis=2) def dstack(xs): """Concatenate variables along third axis (depth wise). Args: xs (list of :class:`~chainer.Variable` or :ref:`ndarray`): Input variables to be concatenated. The variables must have the same ``ndim``. When the variables have the third axis (i.e. :math:`ndim \\geq 3`), the variables must have the same shape along all but the third axis. When the variables do not have the third axis(i.e. :math:`ndim < 3`), the variables must have the same shape. Returns: ~chainer.Variable: Output variable. When the input variables have the third axis (i.e. :math:`ndim \\geq 3`), the shapes of inputs and output are the same along all but the third axis. The length of third axis is the sum of the lengths of inputs' third axis. When the shape of variables are ``(N1, N2)`` (i.e. :math:`ndim = 2`), the shape of output is ``(N1, N2, 2)``. When the shape of variables are ``(N1,)`` (i.e. :math:`ndim = 1`), the shape of output is ``(1, N1, 2)``. When the shape of variables are ``()`` (i.e. :math:`ndim = 0`), the shape of output is ``(1, 1, 2)``. .. admonition:: Example >>> x1 = np.array((1, 2, 3)) >>> x1.shape (3,) >>> x2 = np.array((2, 3, 4)) >>> x2.shape (3,) >>> y = F.dstack((x1, x2)) >>> y.shape (1, 3, 2) >>> y.array array([[[1, 2], [2, 3], [3, 4]]]) >>> x1 = np.arange(0, 6).reshape(3, 2) >>> x1.shape (3, 2) >>> x1 array([[0, 1], [2, 3], [4, 5]]) >>> x2 = np.arange(6, 12).reshape(3, 2) >>> x2.shape (3, 2) >>> x2 array([[ 6, 7], [ 8, 9], [10, 11]]) >>> y = F.dstack([x1, x2]) >>> y.shape (3, 2, 2) >>> y.array array([[[ 0, 6], [ 1, 7]], <BLANKLINE> [[ 2, 8], [ 3, 9]], <BLANKLINE> [[ 4, 10], [ 5, 11]]]) >>> x1 = np.arange(0, 12).reshape(3, 2, 2) >>> x2 = np.arange(12, 18).reshape(3, 2, 1) >>> y = F.dstack([x1, x2]) >>> y.shape (3, 2, 3) >>> y.array array([[[ 0, 1, 12], [ 2, 3, 13]], <BLANKLINE> [[ 4, 5, 14], [ 6, 7, 15]], <BLANKLINE> [[ 8, 9, 16], [10, 11, 17]]]) """ return Dstack().apply((xs))[0]
4,433
30.671429
79
py
chainer
chainer-master/chainer/functions/array/flip.py
import six from chainer import backend from chainer import function_node from chainer.utils import type_check def _flip(array, axis): indices = [slice(None)] * array.ndim indices[axis] = slice(None, None, -1) return array[tuple(indices)] class Flip(function_node.FunctionNode): """Flips an input variable in reverse order along the given axis.""" def __init__(self, axis): if not isinstance(axis, six.integer_types): raise TypeError('axis must be int') self.axis = axis def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type = in_types[0] type_check.expect(x_type.ndim > 0) if self.axis >= 0: type_check.expect(x_type.ndim > self.axis) else: type_check.expect(x_type.ndim >= -self.axis) def forward(self, inputs): xp = backend.get_array_module(*inputs) if hasattr(xp, 'flip'): # numpy.flip is supported from version 1.12.0 return xp.flip(inputs[0], self.axis), else: return _flip(inputs[0], self.axis), def backward(self, indexes, grad_outputs): return flip(grad_outputs[0], self.axis), def flip(x, axis): """Flips an input variable in reverse order along the given axis. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. axis (int): Axis along which the input variable is reversed. Returns: ~chainer.Variable: Output variable. """ return Flip(axis).apply((x,))[0]
1,569
27.035714
78
py
chainer
chainer-master/chainer/functions/array/pad_sequence.py
import numpy import six import chainer from chainer import backend from chainer.backends import cuda from chainer import function_node from chainer import utils from chainer.utils import type_check class PadSequence(function_node.FunctionNode): """Padding arrays to create a matrix.""" def __init__(self, length, padding): self.length = length self.padding = padding def check_type_forward(self, in_types): type_check.expect(in_types.size() > 0) for i, in_type in enumerate(in_types): type_check._argname((in_type,), ('x{}'.format(i),)) type_check.expect( in_type.ndim > 0, in_type.shape[1:] == in_types[0].shape[1:], in_type.dtype == in_types[0].dtype ) if self.length is not None: for in_type in in_types: type_check.expect(in_type.shape[0] <= self.length) def forward(self, xs): xp = backend.get_array_module(*xs) if self.length is None: length = max(len(x) for x in xs) else: length = self.length shape = (len(xs), length) + xs[0].shape[1:] y = xp.empty(shape, xs[0].dtype) if length == 0: return y, # y is an empty array if xp is numpy or any(not x._c_contiguous for x in xs): for i, x in enumerate(xs): l = len(x) if l == length: y[i] = x else: y[i, 0:l] = x y[i, l:] = self.padding else: # This code assumes that all arrays are c_contiguous ptr_shape = (Ellipsis,) + (None,) * xs[0].ndim ptrs = cuda.cupy.array( [x.data for x in xs], numpy.uintp)[ptr_shape] lengths = cuda.cupy.array( [len(x) for x in xs], numpy.int32)[ptr_shape] base = utils.size_of_shape(xs[0].shape[1:]) cuda.elementwise( 'P ptr, int32 length, T pad, int32 base, int32 max_length', 'T y', ''' int d = i / base % max_length; if (d < length) { y = reinterpret_cast<const T*>(ptr)[i % (base * max_length)]; } else { y = pad; } ''', 'pad_sequence_fwd' )(ptrs, lengths, self.padding, base, length, y) return y, def backward(self, indexes, grad_outputs): gy, = grad_outputs inputs = self.inputs if gy.size == 0: # `split` in NumPy 1.9 behaves inconsistently when size is zero. gy = [gy] else: gy = chainer.functions.split_axis(gy, len(inputs), axis=0) return tuple(g[0, :x.shape[0]] for g, x in six.moves.zip(gy, inputs)) def pad_sequence(xs, length=None, padding=0): """Pad given arrays to make a matrix. Args: xs (list of ~chainer.Variable or :ref:`ndarray`): Variables you want to concatenate. length (None or int): Size of the first dimension of a padded array. If it is ``None``, the longest size of the first dimension of ``xs`` is used. padding (int or float): Value to fill. Returns: ~chainer.Variable: A padded matrix. Its shape is ``(n, length, ...)``, where ``n == len(xs)``. """ return PadSequence(length, padding).apply((xs))[0]
3,502
31.435185
79
py
chainer
chainer-master/chainer/functions/array/split_axis.py
import numpy import six import chainer from chainer import backend from chainer.backends import intel64 from chainer import function_node from chainer.utils import collections_abc from chainer.utils import type_check import chainerx _numpy_split_ok = numpy.lib.NumpyVersion(numpy.__version__) >= '1.11.0' def _fix_numpy_split(ys, x, indices_or_sections, axis): """Make the output of np.split compatible with numpy >= 1.11""" if all(y.ndim == x.ndim for y in ys): return ys tmp = [len(t) for t in numpy.split( numpy.empty(x.shape[axis], dtype=numpy.int8), indices_or_sections, 0)] shape = list(x.shape) for i, t in enumerate(tmp): y = ys[i] if y.ndim != x.ndim: assert y.size == 0 shape[axis] = t ys[i] = y.reshape(shape) return ys def _get_indices_or_sections(indices_or_sections): """Checks and convert ``indices_or_sections`` argument Converted value is one of: 1-D numpy.ndarray, list, int, and NumPy int scalar. Returns: A binary tuple in which the 1st element is indices (sequence) and the 2nd element is sections (scalar). Only one of the two is not ``None`` and the other is ``None``. """ ios = indices_or_sections is_seq = False if isinstance(ios, numpy.ndarray): # numpy.ndarray if ios.dtype.kind != 'i' and ios.size > 0: # Note: numpy.array([]) (dtype is float64) should be accepted. raise TypeError('indices_or_sections must be integers') if ios.ndim >= 2: raise TypeError('indices_or_sections must be 1-D sequence') is_seq = ios.ndim != 0 elif isinstance(ios, collections_abc.Sequence): # Any sequence except numpy.ndarray ios = list(ios) is_seq = True elif isinstance(indices_or_sections, six.integer_types): # int pass else: raise TypeError( 'indices_or_sections must be integer or 1-D array.\n' 'Actual: {}'.format(type(indices_or_sections))) if is_seq and chainer.is_debug(): for p, n in six.moves.zip(ios, ios[1:]): if p > n: raise ValueError('indices_or_sections must be sorted') if is_seq: return ios, None else: return None, ios class SplitAxis(function_node.FunctionNode): """Function that splits multiple arrays along the specified axis.""" def __init__(self, indices_or_sections, axis): indices, sections = _get_indices_or_sections(indices_or_sections) assert (indices is None) != (sections is None) self.indices = indices self.sections = sections self.axis = axis def check_type_forward(self, in_types): type_check.expect(in_types.size() == 1) type_check.expect(in_types[0].ndim > self.axis) if self.indices is not None: indices = self.indices if len(indices) > 0: max_index = type_check.make_variable(indices[-1], 'max_index') type_check.expect(in_types[0].shape[self.axis] >= max_index) else: assert self.sections is not None sections = type_check.make_variable(self.sections, 'sections') type_check.expect(in_types[0].shape[self.axis] % sections == 0) @property def indices_or_sections(self): return self.indices if self.indices is not None else self.sections def forward_chainerx(self, inputs): x, = inputs return tuple(chainerx.split(x, self.indices_or_sections, self.axis)) def forward(self, inputs): x, = inputs self._xp = backend.get_array_module(x) # Currently iDeep only supports 4 dims if (intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(inputs, (4,)) and self._ideep_is_supported(inputs)): return self._forward_ideep(inputs) indices_or_sections = self.indices_or_sections ret = self._xp.split(x, indices_or_sections, self.axis) if self._xp == numpy and not _numpy_split_ok: ret = _fix_numpy_split(ret, x, indices_or_sections, self.axis) self._shapes = [r.shape for r in ret] return tuple(ret) def _ideep_is_supported(self, inputs): # Returns True if iDeep supports current configuration of inputs and # arguments. This is workaround for limitation in iDeep internal # implementation. if self.indices is not None: indices = self.indices if len(indices) == 0: return False # Empty sequence if indices[0] == 0: return False # Sequence starting with 0 for i in six.moves.range(1, len(indices)): if indices[i-1] == indices[i]: return False # Sequence with duplicate index else: if self.sections == 1: return False # 1 # Workaround for iDeep segfault issue # See: # https://github.com/chainer/chainer/pull/4281#issuecomment-365830630 # TODO(niboshi): Remove this after iDeep is fixed. # Note: inputs[0].ndim is always 4. if (self.axis == 1 or self.axis == -3) and inputs[0].shape[1] == 8: return False return True def _forward_ideep(self, inputs): x, = inputs offsets = intel64.ideep.intVector() # TODO(iDeep) # bypass python3 issue when transfer array to std::vector<> # https://github.com/SimpleITK/SimpleITK/issues/106 axis = self.axis % x.ndim if self.indices is not None: for i in self.indices: offsets.push_back(int(i)) else: d = x.shape[self.axis] step = d // self.sections for i in six.moves.range(step, d, step): offsets.push_back(i) ret = intel64.ideep.concat.Backward( intel64.ideep.array(x), offsets, axis) self._shapes = [r.shape for r in ret] return ret def backward(self, indexes, grad_outputs): dtype = self.inputs[0].dtype grads = [ self._xp.zeros(shape, dtype=dtype) if gy is None else gy for gy, shape in six.moves.zip(grad_outputs, self._shapes)] return chainer.functions.concat(grads, self.axis), def split_axis(x, indices_or_sections, axis, force_tuple=True): """Splits given variables along an axis. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): A variable to be split. indices_or_sections (int or 1-D array): If this argument is an integer, N, the array will be divided into N equal arrays along axis. If it is a 1-D array of sorted integers, it indicates the positions where the array is split. axis (int): Axis that the input array is split along. force_tuple (bool): If ``True`` (the default) this method returns a tuple even when the number of outputs is one. Otherwise, if ``False`` a Variable will be returned when the number of outputs is one. Returns: tuple or ~chainer.Variable: Tuple of :class:`~chainer.Variable` objects if the number of outputs is more than 1 or :class:`~chainer.Variable` otherwise. When ``force_tuple`` is ``True``, returned value is always a tuple regardless of the number of outputs. """ res = SplitAxis(indices_or_sections, axis).apply((x,)) if force_tuple or len(res) != 1: return res return res[0]
7,645
35.409524
79
py
chainer
chainer-master/chainer/functions/array/transpose_sequence.py
import numpy from chainer import backend from chainer.backends import cuda from chainer import function_node from chainer.utils import type_check def _transpose(xs, length): if length == 0: return () xp = backend.get_array_module(*xs) lengths = numpy.empty(length, dtype=numpy.int32) end = length for i, x in enumerate(xs): len_x = len(x) if len_x == end: continue lengths[len_x:end] = i end = len_x lengths[0:end] = len(xs) if xp is numpy: dtype = xs[0].dtype unit = xs[0].shape[1:] outs = tuple([xp.empty((l,) + unit, dtype=dtype) for l in lengths]) for i, x in enumerate(xs): for p, xi in enumerate(x): outs[p][i] = xi else: offsets1 = numpy.empty(len(xs) + 1, dtype=numpy.int32) offsets1[0] = 0 numpy.cumsum([len(x) for x in xs], out=offsets1[1:]) offsets2 = numpy.empty(length + 1, dtype=numpy.int32) offsets2[0] = 0 numpy.cumsum(lengths, dtype=numpy.int32, out=offsets2[1:]) x = xp.concatenate(xs, axis=0) o = xp.empty_like(x) unit = xs[0].size // len(xs[0]) size = length * len(xs) * unit cuda.elementwise( 'int32 len, int32 unit, raw int32 off1, raw int32 off2, raw T vs', 'raw T hs', ''' int ind = i / unit; int off = i - ind * unit; int y = ind / len; int x = ind - y * len; if (off2[x] + y < off2[x + 1]) { hs[(off2[x] + y) * unit + off] = vs[(off1[y] + x) * unit + off]; } ''', 'transpose_sequence' )(length, unit, cuda.to_gpu(offsets1), cuda.to_gpu(offsets2), x, o, size=size) outs = tuple(xp.split(o, offsets2[1:-1])) return outs class TransposeSequence(function_node.FunctionNode): """Function that transposes a list of Variables.""" def __init__(self, length): self._length = length def check_type_forward(self, xs_type): for p, n in zip(xs_type, xs_type[1:]): type_check.expect( p.shape[0] >= n.shape[0], p.shape[1:] == n.shape[1:], ) def forward(self, xs): if not xs: return () return _transpose(xs, self._length) def backward(self, indexes, grad_outputs): return TransposeSequence(len(self.inputs)).apply(grad_outputs) def transpose_sequence(xs): """Transpose a list of Variables. This function transposes a list of :class:`~chainer.Variable`\\ s and returns a list of :class:`Variable`\\ s. For example a user gives ``[(0, 1, 2, 3), (4, 5), (6)]``, the function returns ``[(0, 4, 6), (1, 5), (2), (3)]``. Note that a given list needs to be sorted by each length of :class:`~chainer.Variable`. Args: xs (list of :class:`~chainer.Variable` or :ref:`ndarray`): Variables to transpose. Returns: tuple of :class:`~chainer.Variable`: Transposed list. .. admonition:: Example >>> lst = [chainer.Variable(np.array([1, 1, 1])), ... chainer.Variable(np.array([2, 2])), ... chainer.Variable(np.array([3]))] >>> lst [variable([1, 1, 1]), variable([2, 2]), variable([3])] >>> transposed = F.transpose_sequence(lst) >>> transposed (variable([1, 2, 3]), variable([1, 2]), variable([1])) """ if not xs: return () return TransposeSequence(len(xs[0])).apply(xs)
3,591
28.68595
78
py
chainer
chainer-master/chainer/functions/array/pad.py
import numpy from chainer import backend from chainer import function_node from chainer.utils import type_check class Pad(function_node.FunctionNode): """Padding of an array.""" def __init__(self, pad_width, mode, **keywords): self.mode = mode self.keywords = keywords self.pad_width = pad_width self.pad_bw = numpy.asarray(pad_width) if self.pad_bw.size == 1: self.pad_bw = numpy.repeat(self.pad_bw, 2) def check_type_forward(self, in_types): # Depending on the arguments, pad_width and keywords, the input value # may be inappropriate. In that case, numpy.pad or cupy.pad will raise # errors, so that only check the size and the dtype in this function. type_check._argname(in_types, ('x',)) x_type = in_types[0] type_check.expect(x_type.dtype.kind == 'f') def forward(self, inputs): xp = backend.get_array_module(*inputs) return xp.pad(inputs[0], self.pad_width, mode=self.mode, **self.keywords), def backward(self, inputs, grad_outputs): gy, = grad_outputs in_shape = self.inputs[0].shape if self.pad_bw.ndim == 1: self.pad_bw = numpy.tile(self.pad_bw, (len(in_shape), 1)) input_idxs = tuple( slice(p[0], p[0] + dim) for dim, p in zip(in_shape, self.pad_bw)) return gy[input_idxs], def pad(x, pad_width, mode, **keywords): """Pad an input variable. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input data. pad_width (int or array-like): Number of values padded to the edges of each axis. mode (str): Specifies how the function fills the periphery of the array. The mode is passed to :func:`numpy.pad` or :func:`cupy.pad`. If it is ``'constant'``, the input is padded by a constant value specified by ``constant_values``. constant_values (int or array-like): Constant values to fill the periphery in the ``'constant'`` mode. Returns: ~chainer.Variable: Output variable. """ return Pad(pad_width, mode, **keywords).apply((x,))[0]
2,221
33.71875
78
py
chainer
chainer-master/chainer/functions/array/resize_images.py
from __future__ import division import numpy from chainer import backend from chainer.backends import cuda from chainer import function_node from chainer.utils import type_check def _infer_lines(B, C, H, W, out_H, out_W, kH, kW): target_size = 2 ** 17 line_size = B * C * (H * W // out_H + kH * kW * out_W) target_lines = target_size // line_size if target_lines < out_H: lines = 1 while True: next_lines = lines * 2 if next_lines > target_lines: break lines = next_lines else: lines = out_H return lines def interpolate_bilinear_cpu(x, v, u, vw, uw): B, C, H, W = x.shape out_H, out_W = v.shape # Interpolation is done by each output panel (i.e. multi lines) # in order to better utilize CPU cache memory. lines = _infer_lines(B, C, H, W, out_H, out_W, 2, 2) vcol = numpy.empty((2, lines, out_W), dtype=v.dtype) ucol = numpy.empty((2, lines, out_W), dtype=u.dtype) wcol = numpy.empty((2, 2, lines, out_W), dtype=x.dtype) y = numpy.empty((B * C, out_H * out_W), dtype=x.dtype) for i in range(0, out_H, lines): l = min(lines, out_H - i) vcol = vcol[:, :l] ucol = ucol[:, :l] wcol = wcol[:, :, :l] i_end = i + l # indices vcol[0] = v[i:i_end] ucol[0] = u[i:i_end] numpy.add(vcol[0], 1, out=vcol[1]) numpy.add(ucol[0], 1, out=ucol[1]) numpy.minimum(vcol[1], H - 1, out=vcol[1]) numpy.minimum(ucol[1], W - 1, out=ucol[1]) # weights # wcol[0, 0] = (1 - uw) * (1 - vw) # wcol[0, 1] = uw * (1 - vw) # wcol[1, 0] = (1 - uw) * vw # wcol[1, 1] = uw * vw wcol[0, 1] = uw[i:i_end] numpy.subtract(1, wcol[0, 1], out=wcol[0, 0]) numpy.multiply(wcol[0], vw[i:i_end], out=wcol[1]) wcol[0] -= wcol[1] # packing to the panel whose shape is (B, C, 2, 2, l, out_W) panel = x[:, :, vcol[:, None], ucol[None, :]] # interpolation panel = panel.reshape((B * C, 4, l * out_W)) weights = wcol.reshape((4, l * out_W)) iout = i * out_W iout_end = i_end * out_W numpy.einsum('ijk,jk->ik', panel, weights, out=y[:, iout:iout_end]) del panel, weights return y.reshape((B, C, out_H, out_W)) def interpolate_bilinear_gpu(x, v, u, vw, uw): B, C, H, W = x.shape out_H, out_W = v.shape y = cuda.cupy.empty((B, C, out_H, out_W), dtype=x.dtype) cuda.elementwise( 'raw T x, S v, S u, T vw, T uw, S H, S W, S outsize', 'T y', ''' // indices S v0 = v; S v1 = min(v + 1, (S)(H - 1)); S u0 = u; S u1 = min(u + 1, (S)(W - 1)); // weights T w0 = (1 - vw) * (1 - uw); T w1 = (1 - vw) * uw; T w2 = vw * (1 - uw); T w3 = vw * uw; // fetch S offset = i / outsize * H * W; T px0 = x[offset + v0 * W + u0]; T px1 = x[offset + v0 * W + u1]; T px2 = x[offset + v1 * W + u0]; T px3 = x[offset + v1 * W + u1]; // interpolate y = (w0 * px0 + w1 * px1) + (w2 * px2 + w3 * px3); ''', 'resize_images_interpolate_bilinear' )(x, v, u, vw, uw, H, W, out_H * out_W, y) return y def interpolate_grad_bilinear_cpu(gy, v, u, vw, uw, H, W): B, C, out_H, out_W = gy.shape # indices vcol = numpy.empty((2, out_H, out_W), dtype=v.dtype) ucol = numpy.empty((2, out_H, out_W), dtype=u.dtype) vcol[0] = v ucol[0] = u numpy.add(vcol[0], 1, out=vcol[1]) numpy.add(ucol[0], 1, out=ucol[1]) numpy.minimum(vcol[1], H - 1, out=vcol[1]) numpy.minimum(ucol[1], W - 1, out=ucol[1]) # weights wcol = numpy.empty((2, 2, out_H, out_W), dtype=gy.dtype) wcol[0, 1] = uw numpy.subtract(1, wcol[0, 1], out=wcol[0, 0]) numpy.multiply(wcol[0], vw, out=wcol[1]) wcol[0] -= wcol[1] # grad gycol = gy.reshape((B * C, 1, 1, out_H, out_W)) * wcol # ravel everything and use `bincount` indices = (vcol[:, None] * W + ucol[None, :]).ravel() offsets = numpy.arange(0, B * C * H * W, H * W, dtype=v.dtype) indices = (offsets[:, None] + indices).ravel() gx = numpy.bincount(indices, weights=gycol.ravel(), minlength=(B * C * H * W)) gx = gx.astype(gy.dtype, copy=False) return gx.reshape((B, C, H, W)) def interpolate_grad_bilinear_gpu(gy, v, u, vw, uw, H, W): B, C, out_H, out_W = gy.shape gx = cuda.cupy.zeros((B * C, H, W), dtype=gy.dtype) cuda.elementwise( 'T gy, S v, S u, T vw, T uw, S H, S W, S outsize', 'raw T gx', ''' // indices S v0 = v; S v1 = min(v + 1, (S)(H - 1)); S u0 = u; S u1 = min(u + 1, (S)(W - 1)); // weights T w0 = (1 - vw) * (1 - uw); T w1 = (1 - vw) * uw; T w2 = vw * (1 - uw); T w3 = vw * uw; // scatter S offset = i / outsize * H * W; atomicAdd(&gx[offset + v0 * W + u0], w0 * gy); atomicAdd(&gx[offset + v0 * W + u1], w1 * gy); atomicAdd(&gx[offset + v1 * W + u0], w2 * gy); atomicAdd(&gx[offset + v1 * W + u1], w3 * gy); ''', 'resize_images_interpolate_grad_bilinear' )(gy, v, u, vw, uw, H, W, out_H * out_W, gx) return gx.reshape((B, C, H, W)) def compute_indices_and_weights(out_size, in_size, mode, align_corners, xp): out_H, out_W = out_size H, W = in_size if mode == 'bilinear': if align_corners: v = xp.linspace(0, H - 1, num=out_H, dtype=numpy.float) u = xp.linspace(0, W - 1, num=out_W, dtype=numpy.float) else: y_scale = H / out_H x_scale = W / out_W v = (xp.arange(out_H, dtype=numpy.float) + 0.5) * y_scale - 0.5 v = xp.maximum(v, 0) u = (xp.arange(out_W, dtype=numpy.float) + 0.5) * x_scale - 0.5 u = xp.maximum(u, 0) vw, v = xp.modf(v) uw, u = xp.modf(u) elif mode == 'nearest': y_scale = H / out_H x_scale = W / out_W v = xp.minimum(xp.floor( xp.arange(out_H, dtype=numpy.float) * y_scale), H - 1) u = xp.minimum(xp.floor( xp.arange(out_W, dtype=numpy.float) * x_scale), W - 1) vw = xp.zeros_like(v) uw = xp.zeros_like(u) return v, u, vw, uw class ResizeImages(function_node.FunctionNode): def __init__(self, output_shape, mode, align_corners): self.out_H = output_shape[0] self.out_W = output_shape[1] assert mode in ['bilinear', 'nearest'] self.mode = mode self.align_corners = align_corners def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type = in_types[0] type_check.expect( x_type.dtype == numpy.float32, x_type.ndim == 4 ) def forward(self, inputs): x, = inputs xp = backend.get_array_module(x) v, u, vw, uw = compute_indices_and_weights( (self.out_H, self.out_W), x.shape[2:], self.mode, self.align_corners, xp) v = v.astype(numpy.intp) u = u.astype(numpy.intp) vw = vw.astype(x.dtype) uw = uw.astype(x.dtype) # Meshgrid-like operation. Meshgrid can cause # performance loss due to memory consumption. # Note that numpy 1.9 doesn't support broadcast_to method. v, u, vw, uw = xp.broadcast_arrays( v[:, None], u[None, :], vw[:, None], uw[None, :]) if xp is numpy: y = interpolate_bilinear_cpu(x, v, u, vw, uw) else: y = interpolate_bilinear_gpu(x, v, u, vw, uw) return y, def backward(self, indexes, grad_outputs): return ResizeImagesGrad( self.inputs[0].shape, (self.out_H, self.out_W), self.mode, self.align_corners).apply(grad_outputs) class ResizeImagesGrad(function_node.FunctionNode): def __init__(self, input_shape, output_shape, mode, align_corners): self.out_H = output_shape[0] self.out_W = output_shape[1] self.input_shape = input_shape assert mode in ['bilinear', 'nearest'] self.mode = mode self.align_corners = align_corners def check_type_forward(self, in_types): type_check._argname(in_types, ('gy',)) gy_type = in_types[0] type_check.expect( gy_type.dtype == numpy.float32, gy_type.ndim == 4 ) def forward(self, inputs): gy, = inputs xp = backend.get_array_module(gy) _, C, H, W = self.input_shape v, u, vw, uw = compute_indices_and_weights( (self.out_H, self.out_W), (H, W), self.mode, self.align_corners, xp) v = v.astype(numpy.intp) u = u.astype(numpy.intp) vw = vw.astype(gy.dtype) uw = uw.astype(gy.dtype) # Meshgrid-like operation. Meshgrid can cause # performance loss due to memory consumption. # Note that numpy 1.9 doesn't support broadcast_to method. v, u, vw, uw = xp.broadcast_arrays( v[:, None], u[None, :], vw[:, None], uw[None, :]) if xp is numpy: gx = interpolate_grad_bilinear_cpu(gy, v, u, vw, uw, H, W) else: gx = interpolate_grad_bilinear_gpu(gy, v, u, vw, uw, H, W) return gx, def backward(self, indexes, grad_outputs): return ResizeImages( (self.out_H, self.out_W), self.mode, self.align_corners).apply(grad_outputs) def resize_images(x, output_shape, *, mode='bilinear', align_corners=True): """Resize images to the given shape. This function resizes 2D data to :obj:`output_shape`. Notation: here is a notation for dimensionalities. - :math:`n` is the batch size. - :math:`c_I` is the number of the input channels. - :math:`h` and :math:`w` are the height and width of the input image, respectively. - :math:`h_O` and :math:`w_O` are the height and width of the output image. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable of shape :math:`(n, c_I, h, w)`. output_shape (tuple): This is a tuple of length 2 whose values are :obj:`(h_O, w_O)`. Note that the order of height and width is opposite of the one in OpenCV. mode ({\'bilinear\', \'nearest\'}): Defines the sampling rule. align_corners (bool): When this value is :obj:`True`, the corners of the input are mapped to the corners of the output. When :obj:`False`, the behavior is the same as OpenCV. Returns: ~chainer.Variable: Resized image whose shape is \ :math:`(n, c_I, h_O, w_O)`. """ return ResizeImages(output_shape, mode, align_corners).apply((x,))[0]
10,949
31.784431
76
py
chainer
chainer-master/chainer/functions/activation/softmax.py
import chainer from chainer import backend from chainer.backends import cuda from chainer import function_node import chainer.functions from chainer.utils import type_check if cuda.cudnn_enabled: cudnn = cuda.cudnn _algorithm = cuda.libcudnn.CUDNN_SOFTMAX_ACCURATE class Softmax(function_node.FunctionNode): """Softmax activation function.""" def __init__(self, axis=1): self.axis = axis def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type, = in_types type_check.expect( x_type.dtype.kind == 'f', -x_type.ndim <= self.axis < x_type.ndim, ) def forward(self, x): xp = backend.get_array_module(*x) if xp is cuda.cupy and chainer.should_use_cudnn('>=auto'): y = cudnn.softmax_forward(x[0], self.axis, _algorithm) else: y = x[0] - x[0].max(axis=self.axis, keepdims=True) xp.exp(y, out=y) y /= y.sum(axis=self.axis, keepdims=True) self.retain_outputs((0,)) return y, def backward(self, indexes, grad_outputs): y = self.get_retained_outputs()[0] gy, = grad_outputs return _SoftmaxGrad(self.axis).apply((y, gy)) class _SoftmaxGrad(function_node.FunctionNode): def __init__(self, axis): self.axis = axis def forward(self, inputs): self.retain_inputs((0, 1)) y, gy = inputs xp = backend.get_array_module(*y) if xp is cuda.cupy and chainer.should_use_cudnn('>=auto'): gx = cudnn.softmax_backward(y, gy, self.axis, _algorithm) else: gx = y * gy sumdx = gx.sum(axis=self.axis, keepdims=True) gx -= y * sumdx return gx, def backward(self, indexes, grad_outputs): y, gy = self.get_retained_inputs() ggx, = grad_outputs gs = chainer.functions.sum(ggx * y, axis=self.axis, keepdims=True) ga = ggx - chainer.functions.broadcast_to(gs, gy.shape) ret = [] if 0 in indexes: s = chainer.functions.broadcast_to(chainer.functions.sum( y * gy, axis=self.axis, keepdims=True), gy.shape) gy2 = ga * gy - ggx * s ret.append(gy2) if 1 in indexes: ggy = ga * y ret.append(ggy) return tuple(ret) def softmax(x, axis=1): """Softmax function. This function computes its softmax along an axis. Let :math:`c = (c_1, c_2, \\dots, c_D)` be the slice of ``x`` along with the axis. For each slice :math:`c`, it computes the function :math:`f(c)` defined as :math:`f(c)={\\exp(c) \\over \\sum_{d} \\exp(c_d)}`. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`n`-dimensional (:math:`n \\geq 2`) float array. axis (int): The axis along which the softmax is to be computed. Returns: ~chainer.Variable: Output variable. A :math:`n`-dimensional (:math:`n \\geq 2`) float array, which is the same shape with x. .. admonition:: Example >>> x = np.array([[0, 1, 2], [0, 2, 4]], np.float32) >>> x array([[0., 1., 2.], [0., 2., 4.]], dtype=float32) >>> y = F.softmax(x, axis=1) >>> y.array array([[0.09003057, 0.24472848, 0.66524094], [0.01587624, 0.11731043, 0.86681336]], dtype=float32) >>> F.sum(y, axis=1).array array([1., 1.], dtype=float32) """ return Softmax(axis=axis).apply((x,))[0]
3,591
29.965517
77
py
chainer
chainer-master/chainer/functions/activation/rrelu.py
import numpy as np import chainer from chainer.backends import cuda from chainer import function_node from chainer.utils import argument from chainer.utils import type_check def _kern(): return cuda.elementwise( 'T cond, T x, T slope', 'T y', 'y = cond >= 0 ? x : (T)(slope * x)', 'rrelu') class RReLU(function_node.FunctionNode): """Randomized Leaky rectifier unit.""" def __init__(self, lower=1. / 8, upper=1. / 3, r=None): if not 0.0 <= lower < 1.0: raise ValueError('lower must be in the range [0, 1)') if not 0.0 <= upper < 1.0: raise ValueError('upper must be in the range [0, 1)') if not lower < upper: raise ValueError('lower must be less than upper') self.lower = lower self.upper = upper self.r = r def check_type_forward(self, in_types): type_check.expect(in_types.size() == 1) x_type, = in_types type_check.expect(x_type.dtype.kind == 'f') if self.r is not None: type_check.expect(x_type.dtype == self.r.dtype) type_check.expect(x_type.shape == self.r.shape) def forward_cpu(self, inputs): x, = inputs if chainer.config.train: if self.r is None: self.r = np.random.uniform( self.lower, self.upper, x.shape ).astype(x.dtype, copy=False) else: self.r = np.full( x.shape, (self.lower + self.upper) / 2, dtype=x.dtype) y = np.where(x >= 0, x, x * self.r) self.retain_outputs((0,)) return y, def forward_gpu(self, inputs): x, = inputs xp = cuda.cupy if chainer.config.train: if self.r is None: self.r = xp.random.uniform( self.lower, self.upper, x.shape ).astype(x.dtype, copy=False) else: self.r = xp.full( x.shape, (self.lower + self.upper) / 2, dtype=x.dtype) y = _kern()(x, x, self.r) self.retain_outputs((0,)) return y, def backward(self, indexes, grad_outputs): y = self.get_retained_outputs()[0].data return _RReLUGrad(y, self.r).apply(grad_outputs) class _RReLUGrad(function_node.FunctionNode): def __init__(self, y, r): self.r = r self.y = y def forward_cpu(self, inputs): gy, = inputs gy = np.where(self.y >= 0, gy, gy * self.r) return gy, def forward_gpu(self, inputs): gy, = inputs gy = _kern()(self.y, gy, self.r) return gy, def backward(self, indexes, grad_outputs): return _RReLUGrad(self.y, self.r).apply(grad_outputs) def rrelu(x, l=1. / 8, u=1. / 3, **kwargs): """rrelu(x, l=1. / 8, u=1. / 3, *, r=None, return_r=False) Randomized Leaky Rectified Liner Unit function. This function is expressed as .. math:: f(x)=\\max(x, rx), where :math:`r` is a random number sampled from a uniform distribution :math:`U(l, u)`. .. note:: The :math:`r` corresponds to :math:`a` in the original paper (https://arxiv.org/pdf/1505.00853.pdf). Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. l (float): The lower bound of the uniform distribution. u (float): The upper bound of the uniform distribution. r (:ref:`ndarray` or None): The r to be used for rrelu. The shape and dtype must be the same as ``x[0]`` and should be on the same device. If ``r`` is not specified or set to ``None``, an ``r`` will be generated randomly according to the given ``l`` and ``u``. If ``r`` is specified, ``l`` and ``u`` will be ignored. return_r (bool): If ``True``, the r used for rrelu is returned altogether with the output variable. The returned ``r`` can latter be reused by passing it to ``r`` argument. Returns: ~chainer.Variable or tuple: When ``return_r`` is ``False`` (default), return the output variable. Otherwise returnes the tuple of the output variable and ``r`` (:ref:`ndarray`). The ``r`` will be on the same device as the input. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. .. admonition:: Example >>> x = np.array([[-1, 0], [2, -3], [-2, 1]], np.float32) >>> x array([[-1., 0.], [ 2., -3.], [-2., 1.]], dtype=float32) >>> F.rrelu(x).array # doctest: +SKIP array([[-0.24850948, 0. ], [ 2. , -0.50844127], [-0.598535 , 1. ]], dtype=float32) """ r = None return_r = False if kwargs: r, return_r = argument.parse_kwargs( kwargs, ('r', r), ('return_r', r), train='train argument is not supported anymore. ' 'Use chainer.using_config') func = RReLU(l, u, r) out, = func.apply((x,)) r = func.r if return_r: return out, r return out
5,240
31.351852
78
py
chainer
chainer-master/chainer/functions/activation/selu.py
from chainer.functions.activation import elu def selu(x, alpha=1.6732632423543772848170429916717, scale=1.0507009873554804934193349852946): """Scaled Exponential Linear Unit function. For parameters :math:`\\alpha` and :math:`\\lambda`, it is expressed as .. math:: f(x) = \\lambda \\left \\{ \\begin{array}{ll} x & {\\rm if}~ x \\ge 0 \\\\ \\alpha (\\exp(x) - 1) & {\\rm if}~ x < 0, \\end{array} \\right. See: https://arxiv.org/abs/1706.02515 Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. alpha (float): Parameter :math:`\\alpha`. scale (float): Parameter :math:`\\lambda`. Returns: ~chainer.Variable: Output variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. """ return scale * elu.elu(x, alpha=alpha)
933
29.129032
78
py
chainer
chainer-master/chainer/functions/activation/softplus.py
import numpy from chainer.backends import cuda from chainer import function_node import chainer.functions from chainer import utils from chainer.utils import type_check class Softplus(function_node.FunctionNode): """Softplus function.""" def __init__(self, beta=1.0): self.beta = float(beta) self.beta_inv = float(1.0 / beta) def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type, = in_types type_check.expect(x_type.dtype.kind == 'f') def forward_cpu(self, inputs): self.retain_inputs((0,)) x = inputs[0] # y = log(1 + exp(beta * x)) / beta bx = self.beta * x y = (numpy.fmax(bx, 0) + numpy.log1p(numpy.exp(-numpy.fabs(bx)))) * self.beta_inv return utils.force_array(y, x.dtype), def forward_gpu(self, inputs): self.retain_inputs((0,)) x = inputs[0] y = cuda.elementwise( 'T x, T beta, T beta_inv', 'T y', ''' T bx = beta * x; y = (max(bx, (T)0) + log1p(exp(-fabs(bx)))) * beta_inv; ''', 'softplus_fwd' )(x, self.beta, self.beta_inv) return y, def backward(self, indexes, grad_outputs): x = self.get_retained_inputs()[0] gy, = grad_outputs return SoftplusGrad((self.beta,)).apply((x, gy)) class SoftplusGrad(function_node.FunctionNode): """Softplus gradient function.""" def __init__(self, inputs): super(SoftplusGrad, self).__init__() self.beta = inputs[0] def check_type_forward(self, in_types): type_check.expect(in_types.size() == 2) x_type, gy_type = in_types type_check.expect(x_type.dtype.kind == 'f') type_check.expect(gy_type.dtype.kind == 'f') def forward_cpu(self, inputs): self.retain_inputs((0, 1)) x, gy = inputs gx = (1 - 1 / (1 + numpy.exp(self.beta * x))) * gy return utils.force_array(gx, x.dtype), def forward_gpu(self, inputs): self.retain_inputs((0, 1)) x, gy = inputs gx = cuda.elementwise( 'T x, T gy, T beta', 'T gx', 'gx = (1 - 1 / (1 + exp(beta * x))) * gy', 'softplus_bwd')(x, gy, self.beta) return gx, def backward(self, indexes, grad_outputs): x, gy = self.get_retained_inputs() ggx, = grad_outputs e = chainer.functions.exp(self.beta * x) gx = ggx * gy * self.beta * e / (1 + e) ** 2 ggy = SoftplusGrad((self.beta,)).apply((x, ggx))[0] return gx, ggy def softplus(x, beta=1.0): """Element-wise softplus function. The softplus function is the smooth approximation of ReLU. .. math:: f(x)=\\frac{1}{\\beta}\\log(1 + \\exp(\\beta x)), where :math:`\\beta` is a parameter. The function becomes curved and akin to ReLU as the :math:`\\beta` is increasing. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. beta (float): Parameter :math:`\\beta`. Returns: ~chainer.Variable: Output variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. .. admonition:: Example >>> x = np.arange(-2, 3, 2).astype(np.float32) >>> x array([-2., 0., 2.], dtype=float32) >>> F.softplus(x, beta=1.0).array array([0.126928 , 0.6931472, 2.126928 ], dtype=float32) """ y, = Softplus(beta=beta).apply((x,)) return y
3,567
28.983193
78
py
chainer
chainer-master/chainer/functions/activation/log_softmax.py
import chainer from chainer import backend from chainer.backends import cuda from chainer import function_node import chainer.functions from chainer.utils import type_check import chainerx if cuda.cudnn_enabled: cudnn = cuda.cudnn _algorithm = cuda.cuda.cudnn.CUDNN_SOFTMAX_LOG # type: ignore def logsumexp(x, axis): xp = backend.get_array_module(x) m = x.max(axis=axis, keepdims=True) y = x - m xp.exp(y, out=y) s = y.sum(axis=axis, keepdims=True) xp.log(s, out=s) m += s return m def _log_softmax(x, axis=1): if chainer.should_use_cudnn('>=auto'): xp = backend.get_array_module(x) if xp is cuda.cupy: return cudnn.softmax_forward(x, axis, _algorithm) log_z = logsumexp(x, axis) y = x - log_z return y class LogSoftmax(function_node.FunctionNode): """Log-softmax activation function.""" def __init__(self, axis=1): self.axis = axis def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type, = in_types type_check.expect( x_type.dtype.kind == 'f', -x_type.ndim <= self.axis < x_type.ndim, ) def forward_chainerx(self, xs): return chainerx.log_softmax(xs[0], axis=self.axis), def forward(self, xs): y = _log_softmax(xs[0], axis=self.axis) self._x_xp = backend.get_array_module(*xs) self._x_shape = xs[0].shape self._x_dtype = xs[0].dtype self.retain_outputs((0,)) return y, def backward(self, indexes, gy): y = self.get_retained_outputs()[0] return LogSoftmaxGrad( self._x_xp, self._x_shape, self._x_dtype, self.axis).apply( (y, gy[0])) class LogSoftmaxGrad(function_node.FunctionNode): def __init__(self, x_xp, x_shape, x_dtype, axis): self._x_xp = x_xp self._x_shape = x_shape self._x_dtype = x_dtype self.axis = axis def forward(self, inputs): self.retain_inputs((0, 1)) y, gy = inputs xp = self._x_xp if xp is cuda.cupy and chainer.should_use_cudnn('>=auto'): gx = cudnn.softmax_backward(y, gy, self.axis, _algorithm) else: gx = gy - xp.exp(y) * gy.sum(axis=self.axis, keepdims=True) return gx, def backward(self, indexes, ggx): y, gy = self.get_retained_inputs() ret = [] exp_y = chainer.functions.exp(y) if 0 in indexes: gy_sum = chainer.functions.sum(gy, self.axis, True) gy_sum = chainer.functions.broadcast_to(gy_sum, gy.shape) g0 = -ggx[0] * exp_y * gy_sum ret.append(g0) if 1 in indexes: # TODO(Kenta Oono): implement it with double-backpropable F.matmul a = chainer.functions.sum(ggx[0] * exp_y, self.axis, True) a = chainer.functions.broadcast_to(a, gy.shape) g1 = ggx[0] - a ret.append(g1) return ret def log_softmax(x, axis=1): """Channel-wise log-softmax function. This function computes its logarithm of softmax along the second axis. Let :math:`c = (c_1, c_2, \\dots, c_D)` be the slice of ``x`` along with the second axis. For each slice :math:`c`, it computes the logarithm of the function :math:`f(c)` defined as .. math:: f(c) = {\\exp(c) \\over \\sum_{d} \\exp(c_d)}. This method is theoretically equivalent to ``log(softmax(x))`` but is more stable. .. note:: ``log(softmax(x))`` may cause underflow when ``x`` is too small, because ``softmax(x)`` may returns ``0``. ``log_softmax`` method is more stable. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`n`-dimensional (:math:`n \\geq 2`) float array. axis (int): The axis along which the softmax is to be computed. Returns: ~chainer.Variable: Output variable. A :math:`n`-dimensional (:math:`n \\geq 2`) float array, which is the same shape with x. .. seealso:: :func:`~chainer.functions.softmax` .. admonition:: Example >>> x = np.array([[0, 1, 2], [0, 2, 4]], np.float32) >>> x array([[0., 1., 2.], [0., 2., 4.]], dtype=float32) >>> F.log_softmax(x).array array([[-2.407606 , -1.4076059 , -0.4076059 ], [-4.1429315 , -2.1429315 , -0.14293146]], dtype=float32) >>> np.allclose(F.log_softmax(x).data, F.log(F.softmax(x)).data) True """ return LogSoftmax(axis=axis).apply((x,))[0]
4,633
29.688742
78
py
chainer
chainer-master/chainer/functions/activation/maxout.py
from chainer.functions.array import reshape from chainer.functions.math import minmax from chainer.utils import type_check def maxout(x, pool_size, axis=1): """Maxout activation function. It accepts an input tensor ``x``, reshapes the ``axis`` dimension (say the size being ``M * pool_size``) into two dimensions ``(M, pool_size)``, and takes maximum along the ``axis`` dimension. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`n`-dimensional (:math:`n \\ge` ``axis``) float array. In general, its first dimension is assumed to be the *minibatch dimension*. The other dimensions are treated as one concatenated dimension. pool_size (int): The size used for downsampling of pooling layer. axis (int): The ``axis`` dimension to be reshaped. The size of ``axis`` dimension should be ``M * pool_size``. Returns: ~chainer.Variable: Output variable. The shape of the output is same as ``x`` except that ``axis`` dimension is transformed from ``M * pool_size`` to ``M``. .. seealso:: :class:`~chainer.links.Maxout` .. admonition:: Example Typically, ``x`` is the output of a linear layer or a convolution layer. The following is the example where we use :func:`maxout` in combination with a Linear link. >>> in_size, out_size, pool_size = 10, 10, 10 >>> bias = np.arange(out_size * pool_size).astype(np.float32) >>> l = L.Linear(in_size, out_size * pool_size, initial_bias=bias) >>> x = np.zeros((1, in_size), np.float32) # prepare data >>> x = l(x) >>> y = F.maxout(x, pool_size) >>> x.shape (1, 100) >>> y.shape (1, 10) >>> x.reshape((out_size, pool_size)).array array([[ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9.], [10., 11., 12., 13., 14., 15., 16., 17., 18., 19.], [20., 21., 22., 23., 24., 25., 26., 27., 28., 29.], [30., 31., 32., 33., 34., 35., 36., 37., 38., 39.], [40., 41., 42., 43., 44., 45., 46., 47., 48., 49.], [50., 51., 52., 53., 54., 55., 56., 57., 58., 59.], [60., 61., 62., 63., 64., 65., 66., 67., 68., 69.], [70., 71., 72., 73., 74., 75., 76., 77., 78., 79.], [80., 81., 82., 83., 84., 85., 86., 87., 88., 89.], [90., 91., 92., 93., 94., 95., 96., 97., 98., 99.]], \ dtype=float32) >>> y.array array([[ 9., 19., 29., 39., 49., 59., 69., 79., 89., 99.]], \ dtype=float32) """ if pool_size <= 0: raise ValueError('pool_size must be a positive integer.') x_shape = x.shape if x_shape[axis] % pool_size != 0: expect = 'x.shape[axis] % pool_size == 0' actual = 'x.shape[axis]={}, pool_size={}'.format( x_shape[axis], pool_size) msg = 'axis dimension must be divided by pool_size' raise type_check.InvalidType(expect, actual, msg) shape = (x_shape[:axis] + (x_shape[axis] // pool_size, pool_size) + x_shape[axis + 1:]) x = reshape.reshape(x, shape) return minmax.max(x, axis=axis + 1)
3,309
38.879518
77
py
chainer
chainer-master/chainer/functions/activation/leaky_relu.py
from chainer.backends import cuda from chainer.backends import intel64 from chainer import function_node from chainer.utils import type_check _kern = None def _get_kern(): global _kern if _kern is None: _kern = cuda.elementwise( 'T cond, T x, T slope', 'T y', 'y = cond <= 0 ? (T)(slope * x) : x', 'lrelu') return _kern class LeakyReLU(function_node.FunctionNode): """Leaky rectifier unit.""" def __init__(self, slope=0.2): self.slope = slope def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type, = in_types type_check.expect(x_type.dtype.kind == 'f') def forward_cpu(self, inputs): if (intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(inputs)): return self.forward_ideep(inputs) x, = inputs y = x.copy() y[x <= 0] *= self.slope if self.slope >= 0: self.retain_outputs((0,)) else: self.retain_inputs((0,)) return y, def forward_ideep(self, inputs): x, = inputs y = intel64.ideep.relu.Forward( intel64.ideep.array(x), self.slope) if self.slope >= 0: self.retain_outputs((0,)) else: self.retain_inputs((0,)) return y, def forward_gpu(self, inputs): x, = inputs y = _get_kern()(x, x, self.slope) if self.slope >= 0: self.retain_outputs((0,)) else: self.retain_inputs((0,)) return y, def backward(self, indexes, grad_outputs): if self.slope >= 0: cond = self.get_retained_outputs()[0].array else: cond = self.get_retained_inputs()[0].array return _LeakyReLUGrad(cond, self.slope).apply(grad_outputs) class _LeakyReLUGrad(function_node.FunctionNode): def __init__(self, cond, slope): self.cond = cond self.slope = slope def forward_cpu(self, inputs): if (intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(inputs)): return self.forward_ideep(inputs) gy, = inputs gy = gy.copy() gy[self.cond <= 0] *= self.slope return gy, def forward_ideep(self, inputs): gy, = inputs gy = intel64.ideep.relu.Backward( intel64.ideep.array(self.cond), intel64.ideep.array(gy), self.slope) return gy, def forward_gpu(self, inputs): gy, = inputs gy = _get_kern()(self.cond, gy, self.slope) return gy, def backward(self, indexes, grad_outputs): return _LeakyReLUGrad(self.cond, self.slope).apply(grad_outputs) def leaky_relu(x, slope=0.2): """Leaky Rectified Linear Unit function. This function is expressed as .. math:: f(x) = \\left \\{ \\begin{array}{ll} x & {\\rm if}~ x \\ge 0 \\\\ ax & {\\rm if}~ x < 0, \\end{array} \\right. where :math:`a` is a configurable slope value. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. slope (float): Slope value :math:`a`. Returns: ~chainer.Variable: Output variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. .. admonition:: Example >>> x = np.array([[-1, 0], [2, -3], [-2, 1]], np.float32) >>> x array([[-1., 0.], [ 2., -3.], [-2., 1.]], dtype=float32) >>> F.leaky_relu(x, slope=0.2).array array([[-0.2, 0. ], [ 2. , -0.6], [-0.4, 1. ]], dtype=float32) """ return LeakyReLU(slope).apply((x,))[0]
3,792
25.900709
78
py
chainer
chainer-master/chainer/functions/activation/hard_sigmoid.py
import numpy from chainer.backends import cuda from chainer import function_node from chainer import utils from chainer.utils import type_check class HardSigmoid(function_node.FunctionNode): """Hard-sigmoid function.""" def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type, = in_types type_check.expect(x_type.dtype.kind == 'f') def forward_cpu(self, inputs): x = inputs[0] y = numpy.clip(x * 0.2 + 0.5, 0.0, 1.0) self.retain_inputs((0,)) return utils.force_array(y, x.dtype), def forward_gpu(self, inputs): x = inputs[0] self.retain_inputs((0,)) return cuda.elementwise( 'T x', 'T y', 'y = min(1.0, max(0.0, x * 0.2 + 0.5))', 'hard_sigmoid_fwd' )(x), def backward(self, indexes, grad_outputs): x, = self.get_retained_inputs() return HardSigmoidGrad(x.data).apply(grad_outputs) class HardSigmoidGrad(function_node.FunctionNode): """Hard-sigmoid gradient function.""" def __init__(self, x): self.x = x def check_type_forward(self, in_types): type_check._argname(in_types, ('gy',)) type_check.expect( in_types[0].dtype.kind == 'f', in_types[0].dtype == self.x.dtype ) def forward_cpu(self, inputs): gy, = inputs gx = ((-2.5 < self.x) & (self.x < 2.5)) * gy * 0.2 return utils.force_array(gx, self.x.dtype), def forward_gpu(self, inputs): gy, = inputs return cuda.elementwise( 'T x, T g', 'T gx', 'gx = fabs(x) < 2.5 ? 0.2 * g : 0', 'hard_sigmoid_bwd' )(self.x, gy), def backward(self, indexes, grad_outputs): return HardSigmoidGrad(self.x).apply(grad_outputs) def hard_sigmoid(x): """Element-wise hard-sigmoid function. This function is defined as .. math:: f(x) = \\left \\{ \\begin{array}{ll} 0 & {\\rm if}~ x < -2.5 \\\\ 0.2 x + 0.5 & {\\rm if}~ -2.5 < x < 2.5 \\\\ 1 & {\\rm if}~ 2.5 < x. \\end{array} \\right. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. Returns: ~chainer.Variable: Output variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. .. admonition:: Example It maps the input values into the range of :math:`[0, 1]`. >>> x = np.array([-2.6, -1, 0, 1, 2.6]) >>> x array([-2.6, -1. , 0. , 1. , 2.6]) >>> F.hard_sigmoid(x).array array([0. , 0.3, 0.5, 0.7, 1. ]) """ return HardSigmoid().apply((x,))[0]
2,749
25.442308
78
py
chainer
chainer-master/chainer/functions/activation/elu.py
import numpy from chainer.backends import cuda from chainer import function_node from chainer import utils from chainer.utils import type_check class ELU(function_node.FunctionNode): """Exponential Linear Unit.""" def __init__(self, alpha=1.0): self.alpha = float(alpha) def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type, = in_types type_check.expect(x_type.dtype.kind == 'f') def forward_cpu(self, inputs): if self.alpha < 0: self.retain_inputs((0,)) x, = inputs y = x.copy() negzero_indices = y <= 0 y[negzero_indices] = self.alpha * numpy.expm1(y[negzero_indices]) self.retain_outputs((0,)) return y, def forward_gpu(self, inputs): if self.alpha < 0: self.retain_inputs((0,)) x, = inputs y = cuda.elementwise( 'T x, T alpha', 'T y', 'y = x > 0 ? x : (T)(alpha * expm1(x))', 'elu_fwd')(x, self.alpha) self.retain_outputs((0,)) return y, def backward(self, indexes, grad_outputs): y, = self.get_retained_outputs() if self.alpha < 0: cond, = self.get_retained_inputs() else: cond = y gy, = grad_outputs return ELUGrad(self.alpha, cond.array).apply((y,))[0] * gy, class ELUGrad(function_node.FunctionNode): """Exponential Linear Unit gradient function.""" def __init__(self, alpha, cond): self.alpha = alpha self.cond = cond def forward_cpu(self, inputs): y, = inputs gx = utils.force_array(y + y.dtype.type(self.alpha)) gx[self.cond > 0] = 1 return gx, def forward_gpu(self, inputs): y, = inputs gx = cuda.elementwise( 'T y, T alpha, T cond', 'T gx', 'gx = cond > 0 ? (T)1 : (T)(y + alpha)', 'elu_bwd')(y, self.alpha, self.cond) return gx, def backward(self, indexes, grad_outputs): ggx, = grad_outputs gy2 = ggx * (self.cond <= 0) return gy2, def elu(x, alpha=1.0): """Exponential Linear Unit function. For a parameter :math:`\\alpha`, it is expressed as .. math:: f(x) = \\left \\{ \\begin{array}{ll} x & {\\rm if}~ x \\ge 0 \\\\ \\alpha (\\exp(x) - 1) & {\\rm if}~ x < 0, \\end{array} \\right. See: https://arxiv.org/abs/1511.07289 Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. alpha (float): Parameter :math:`\\alpha`. Default is 1.0. Returns: ~chainer.Variable: Output variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. .. admonition:: Example >>> x = np.array([[-1, 0], [2, -3]], np.float32) >>> x array([[-1., 0.], [ 2., -3.]], dtype=float32) >>> y = F.elu(x, alpha=1.) >>> y.array array([[-0.63212055, 0. ], [ 2. , -0.95021296]], dtype=float32) """ return ELU(alpha=alpha).apply((x,))[0]
3,179
26.413793
78
py
chainer
chainer-master/chainer/functions/activation/swish.py
import numpy import chainer from chainer.backends import cuda from chainer import function_node from chainer import utils from chainer.utils import type_check def _get_extended_shape(beta, x): return (1,) + beta.shape + (1,) * (x.ndim - beta.ndim - 1) def _get_reduction_axes(beta, x): return (0,) + tuple(range(1 + beta.ndim, x.ndim)) def _sigmoid(x): half = x.dtype.type(0.5) return numpy.tanh(x * half) * half + half _preamble = ''' template <typename T> __device__ T sigmoid(T x) { const T half = 0.5; return tanh(x * half) * half + half; } ''' class Swish(function_node.FunctionNode): def check_type_forward(self, in_types): type_check._argname(in_types, ('x', 'beta')) x_type, beta_type = in_types type_check.expect( x_type.dtype.kind == 'f', beta_type.dtype == x_type.dtype, beta_type.ndim <= x_type.ndim - 1, beta_type.shape == x_type.shape[1:1 + type_check.eval(beta_type.ndim)] ) def forward_cpu(self, inputs): self.retain_inputs((0, 1)) x, beta = inputs beta = beta.reshape(_get_extended_shape(beta, x)) y = x * _sigmoid(beta * x) return y, def forward_gpu(self, inputs): self.retain_inputs((0, 1)) x, beta = inputs beta = beta.reshape(_get_extended_shape(beta, x)) # Eliminating intermediate variable `bx` somehow degrades the # precision. y = cuda.elementwise( 'T x, T beta', 'T y', ''' T bx = beta * x; y = x * sigmoid(bx); ''', 'swish_fwd', preamble=_preamble )(x, beta) return y, def backward(self, indexes, grad_outputs): x, beta = self.get_retained_inputs() gy, = grad_outputs shape = _get_extended_shape(beta, x) reduction_axes = _get_reduction_axes(beta, x) return SwishGrad(shape, reduction_axes).apply((x, beta, gy)) class SwishGrad(function_node.FunctionNode): def __init__(self, extended_shape, reduction_axes): super(SwishGrad, self).__init__() self.extended_shape = extended_shape self.reduction_axes = reduction_axes def forward_cpu(self, inputs): self.retain_inputs((0, 1, 2)) x, beta, gy = inputs beta = beta.reshape(self.extended_shape) sig = _sigmoid(beta * x) y = x * sig by = beta * y one = x.dtype.type(1) gx = gy * (by + sig * (one - by)) gb = gy * y * (x - y) gb = utils.force_array(gb.sum(axis=self.reduction_axes)) return gx, gb def forward_gpu(self, inputs): self.retain_inputs((0, 1, 2)) x, beta, gy = inputs beta = beta.reshape(self.extended_shape) gx, gb = cuda.elementwise( 'T x, T beta, T gy', 'T gx, T gb', ''' T bx = beta * x; T sig = sigmoid(bx); T y = x * sig; T by = beta * y; gx = gy * (by + sig * (1 - by)); gb = gy * y * (x - y); ''', 'swish_bwd', preamble=_preamble )(x, beta, gy) gb = utils.force_array(gb.sum(axis=self.reduction_axes)) return gx, gb def backward(self, indexes, grad_outputs): x, beta, gy = self.get_retained_inputs() beta = chainer.functions.broadcast_to( beta.reshape(self.extended_shape), gy.shape) ggx, ggb = grad_outputs ggb = chainer.functions.broadcast_to( ggb.reshape(self.extended_shape), gy.shape) sig = chainer.functions.sigmoid(beta * x) y = x * sig by = beta * y one_minus_sig = 1 - sig sig_one_minus_by = sig * (1 - by) y_x_minus_y = y * (x - y) x_minus_2y = x - 2 * y ret = [] common = gy * y * (2 + beta * x_minus_2y) * one_minus_sig if 0 in indexes: gx = ggx * gy * beta * one_minus_sig * \ (by + 2 * sig_one_minus_by) + ggb * common ret.append(chainer.functions.cast(gx, x.dtype)) if 1 in indexes: gb = ggx * common + ggb * gy * y_x_minus_y * x_minus_2y gb = chainer.functions.sum(gb, axis=self.reduction_axes) ret.append(chainer.functions.cast(gb, beta.dtype)) if 2 in indexes: ggy = ggx * (by + sig_one_minus_by) + ggb * y_x_minus_y ret.append(chainer.functions.cast(ggy, gy.dtype)) return ret def swish(x, beta): """Swish activation function. .. math:: f(x, \\beta) = x \\cdot \\sigma(\\beta x), where :math:`\\sigma(\\cdot)` is the sigmoid function. It has the following properties: .. math:: f(x, 0) &= \\frac{x}{2}, \\\\ \\lim_{\\beta \\to \\infty} f(x, \\beta) &= \\max(0, x). Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable of shape :math:`(s_B, s_1, s_2, ..., s_N)`, where :math:`s_B` is assumed to be the *minibatch dimension*. beta (:class:`~chainer.Variable` or :ref:`ndarray`): Parameter variable :math:`\\beta` of shape :math:`(s_1, s_2, ..., s_M)`, where :math:`M` is an arbitrary integer between :math:`0 \\leq M \\leq N`. The number of dimensions of ``beta`` will be matched with ``x`` by reshaping it as :math:`(1, s_1, ..., s_M, 1, ... 1)`, then ``beta`` and ``x`` are multiplied together in an element-wise manner. Returns: ~chainer.Variable: Output variable of the same shape as ``x``. .. warning:: :math:`\\beta` is a trainable parameter in the original paper (https://arxiv.org/abs/1710.05941). To train :math:`\\beta`, use :class:`chainer.links.Swish` instead. .. seealso:: :class:`chainer.links.Swish` to manage the model parameter ``beta``. """ y, = Swish().apply((x, beta)) return y
6,024
30.217617
79
py
chainer
chainer-master/chainer/functions/activation/__init__.py
0
0
0
py
chainer
chainer-master/chainer/functions/activation/tanh.py
import numpy import chainer from chainer.backends import cuda from chainer import function_node from chainer import utils from chainer.utils import type_check import chainerx if cuda.cudnn_enabled: cudnn = cuda.cudnn _mode = cuda.libcudnn.CUDNN_ACTIVATION_TANH class Tanh(function_node.FunctionNode): """Hyperbolic tangent function.""" def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) type_check.expect(in_types[0].dtype.kind == 'f') def forward_chainerx(self, x): return chainerx.tanh(x[0]), def forward_cpu(self, x): y = utils.force_array(numpy.tanh(x[0])) self.retain_outputs((0,)) self._use_cudnn = False return y, def forward_gpu(self, x): if chainer.should_use_cudnn('==always') and x[0].flags.c_contiguous: y = cudnn.activation_forward(x[0], _mode) self.retain_inputs((0,)) self._use_cudnn = True else: y = cuda.cupy.empty_like(x[0]) cuda.cupy.tanh(x[0], out=y) self._use_cudnn = False self.retain_outputs((0,)) return y, def backward(self, indexes, grad_outputs): if self._use_cudnn: x = self.get_retained_inputs()[0].data else: x = None y = self.get_retained_outputs()[0] gy = grad_outputs[0] return TanhGrad(x).apply((y, gy)) class TanhGrad(function_node.FunctionNode): def __init__(self, x): super(TanhGrad, self).__init__() # The original input `x` is only required for cuDNN. # If it is None, this class does not use cuDNN. # Note that x must be c-contiguous and it is checked # in Tanh.forward_gpu. self.x = x def forward_cpu(self, inputs): self.retain_inputs((0, 1)) y, gy = inputs one = y.dtype.type(1) return utils.force_array(gy * (one - y * y)), def forward_gpu(self, inputs): self.retain_inputs((0, 1)) y, gy = inputs if (chainer.should_use_cudnn('==always') and self.x is not None and gy.flags.c_contiguous): gx = cudnn.activation_backward(self.x, y, gy, _mode) else: gx = cuda.elementwise( 'T y, T gy', 'T gx', 'gx = gy * (1 - y * y)', 'tanh_bwd')(y, gy) return gx, def backward(self, indexes, grad_outputs): y, gy = self.get_retained_inputs() ggx = grad_outputs[0] y_mul_ggx = y * ggx grad_y = -2 * gy * y_mul_ggx ggy = ggx - y * y_mul_ggx return grad_y, ggy def tanh(x): """Elementwise hyperbolic tangent function. .. math:: f(x)=\\tanh(x). Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. Returns: ~chainer.Variable: Output variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. .. admonition:: Example >>> x = np.arange(-1, 4, 2).astype(np.float32) >>> x array([-1., 1., 3.], dtype=float32) >>> F.tanh(x).array array([-0.7615942, 0.7615942, 0.9950548], dtype=float32) """ return Tanh().apply((x,))[0]
3,299
27.205128
78
py
chainer
chainer-master/chainer/functions/activation/crelu.py
import six import chainer from chainer import backend from chainer import function_node from chainer.utils import type_check class CReLU(function_node.FunctionNode): """Concatenated Rectified Linear Unit.""" def __init__(self, axis=1): if not isinstance(axis, six.integer_types): raise TypeError('axis must be an integer value') self.axis = axis def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) type_check.expect( in_types[0].dtype.kind == 'f', in_types[0].ndim > self.axis, in_types[0].ndim >= -self.axis ) def get_output_shape(self, input_shape): output_shape = list(input_shape) output_shape[self.axis] *= 2 return tuple(output_shape) def forward(self, inputs): x, = inputs xp = backend.get_array_module(x) y = xp.empty(self.get_output_shape(x.shape), dtype=x.dtype) y_former, y_latter = xp.split(y, 2, axis=self.axis) zero = x.dtype.type(0) xp.maximum(zero, x, out=y_former) xp.maximum(zero, -x, out=y_latter) self.retain_inputs((0,)) return y, def backward(self, indexes, grad_outputs): x, = self.get_retained_inputs() gy, = grad_outputs gy_former, gy_latter = chainer.functions.split_axis( gy, 2, axis=self.axis) return gy_former * (x.data > 0) - gy_latter * (x.data < 0), def crelu(x, axis=1): """Concatenated Rectified Linear Unit function. This function is expressed as follows .. math:: f(x) = (\\max(0, x), \\max(0, -x)). Here, two output values are concatenated along an axis. See: https://arxiv.org/abs/1603.05201 Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. axis (int): Axis that the output values are concatenated along. Default is 1. Returns: ~chainer.Variable: Output variable of concatenated array. If the axis is 1, A :math:`(s_1, s_2 \\times 2, ..., s_N)`-shaped float array. .. admonition:: Example >>> x = np.array([[-1, 0], [2, -3]], np.float32) >>> x array([[-1., 0.], [ 2., -3.]], dtype=float32) >>> y = F.crelu(x, axis=1) >>> y.array array([[0., 0., 1., 0.], [2., 0., 0., 3.]], dtype=float32) """ return CReLU(axis=axis).apply((x,))[0]
2,525
28.717647
79
py
chainer
chainer-master/chainer/functions/activation/relu.py
import numpy import chainer from chainer.backends import cuda from chainer.backends import intel64 from chainer import function_node from chainer import utils from chainer.utils import type_check import chainerx if cuda.available: _relu_grad2_kernel = cuda.elementwise( 'T y, T gy', 'T gx', 'gx = y > 0 ? gy : (T)0', 'relu_bwd') if cuda.cudnn_enabled: cudnn = cuda.cudnn _mode = cuda.cuda.cudnn.CUDNN_ACTIVATION_RELU # type: ignore class ReLU(function_node.FunctionNode): """Rectified Linear Unit.""" is_elementwise = True _use_cudnn = False def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) type_check.expect(in_types[0].dtype.kind == 'f') def forward_chainerx(self, inputs): x, = inputs return chainerx.maximum(x, 0), def forward_cpu(self, inputs): if (intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(inputs)): return self.forward_ideep(inputs) x, = inputs y = numpy.maximum(x, 0, dtype=x.dtype) self.retain_outputs((0,)) return utils.force_array(y), def forward_ideep(self, inputs): x, = inputs y = intel64.ideep.relu.Forward(intel64.ideep.array(x)) self.retain_outputs((0,)) return y, def forward_gpu(self, inputs): x, = inputs if chainer.should_use_cudnn('>=auto') and x.flags.c_contiguous: self._use_cudnn = True y = cudnn.activation_forward(x, _mode) else: y = cuda.cupy.maximum(x, 0, dtype=x.dtype) self.retain_outputs((0,)) return y, def backward(self, indexes, grad_outputs): gy, = grad_outputs y, = self.get_retained_outputs() y_arr = y.raw_array if self._use_cudnn and chainer.should_use_cudnn('>=auto'): # cuDNN implementation return ReLUGradCudnn(y_arr).apply((gy,)) # Generic implementation return ReLUGrad2(y_arr).apply((gy,)) class ReLUGrad2(function_node.FunctionNode): """Computes the gradient of the ReLU function. This function takes 2 variables b and c, and computes f(b, c) = sign(b) * c with backpropagation where operations are done in elementwise manner and sign(x) = 1 when x > 0 is positive and 0 otherwise. As the gradient of f with respect to b is 0, we do not backpropagate errors toward b for computational efficiency. """ def __init__(self, b): super(ReLUGrad2, self).__init__() self.b = b def forward_cpu(self, inputs): if (intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(inputs)): return self.forward_ideep(inputs) gy, = inputs gx = gy * (self.b > 0) return utils.force_array(gx, dtype=gy.dtype), def forward_ideep(self, inputs): gy, = inputs gx = intel64.ideep.relu.Backward( intel64.ideep.array(self.b), intel64.ideep.array(gy)) return gx, def forward_gpu(self, inputs): gx = _relu_grad2_kernel(self.b, inputs[0]) return gx, def backward(self, indexes, grad_outputs): return ReLUGrad2(self.b).apply(grad_outputs) class ReLUGradCudnn(function_node.FunctionNode): """Computes the gradient of the ReLU function. This function takes 3 variables a, b, and c, and computes f(a, b, c) = sign(b) * c with backpropagation where operations are dones in elementwise manner and sign(x) = 1 if x > 0 is positive and 0 otherwise. As the gradient of f with respect to a and b are 0, we do not backpropagate errors toward them for computational efficiency. """ is_elementwise = True def __init__(self, y): super(ReLUGradCudnn, self).__init__() self.y = y def forward(self, inputs): gy, = inputs return cudnn.activation_backward(self.y, self.y, gy, _mode), def backward(self, indexes, grad_outputs): return ReLUGrad2(self.y).apply(grad_outputs) def relu(x): """Rectified Linear Unit function. .. math:: f(x)=\\max(0, x). Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. Returns: ~chainer.Variable: Output variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. .. admonition:: Example >>> x = np.array([[-1, 0], [2, -3], [-2, 1]], np.float32) >>> np.any(x < 0) True >>> y = F.relu(x) >>> np.any(y.array < 0) False >>> y.shape (3, 2) """ y, = ReLU().apply((x,)) return y
4,736
27.196429
78
py
chainer
chainer-master/chainer/functions/activation/prelu.py
import numpy import chainer from chainer.backends import cuda from chainer import function_node from chainer.utils import type_check def _fwd_kern(): return cuda.elementwise( 'T x, T cond, T W', 'T y', 'y = cond >= 0 ? x : (T)(x * W)', 'prelu') def _get_extended_shape(W, x): return (1,) + W.shape + (1,) * (x.ndim - W.ndim - 1) def _get_reduce_axes(W, x): return (0,) + tuple(range(1 + W.ndim, x.ndim)) class PReLUFunction(function_node.FunctionNode): """Parametric Rectified Linear Unit function.""" def check_type_forward(self, in_types): type_check._argname(in_types, ('x', 'W')) x_type, W_type = in_types type_check.expect( x_type.dtype.kind == 'f', W_type.dtype == x_type.dtype, x_type.ndim >= W_type.ndim + 1, x_type.shape[1:1 + type_check.eval(W_type.ndim)] == W_type.shape ) def forward_cpu(self, inputs): x, W = inputs y = x.copy() masked = numpy.ma.masked_greater_equal(y, 0, copy=False) shape = _get_extended_shape(W, y) masked *= W.reshape(shape) self.retain_inputs((0, 1)) return y, def forward_gpu(self, inputs): x, W = inputs shape = _get_extended_shape(W, x) y = _fwd_kern()(x, x, W.reshape(shape)) self.retain_inputs((0, 1)) return y, def backward(self, indexes, grad_outputs): x, W = self.get_retained_inputs() gy, = grad_outputs return PReLUFunctionGrad( x.data, _get_reduce_axes(W, x), _get_extended_shape(W, x)).apply((x, W, gy)) class PReLUFunctionGrad(function_node.FunctionNode): """Parametric Rectified Linear Unit gradient function.""" def __init__(self, cond, reduce_axes, extended_shape): self.cond = cond self.reduce_axes = reduce_axes self.extended_shape = extended_shape def check_type_forward(self, in_types): type_check._argname(in_types, ('x', 'W', 'gy')) x_type, W_type, gy_type = in_types type_check.expect( x_type.dtype.kind == 'f', W_type.dtype == x_type.dtype, gy_type.dtype == x_type.dtype, x_type.ndim >= W_type.ndim + 1, x_type.shape[1:1 + type_check.eval(W_type.ndim)] == W_type.shape, gy_type.shape == x_type.shape ) def forward_cpu(self, inputs): x, W, gy = inputs mask = self.cond >= 0 masked = numpy.where(mask, 0, x * gy) if self.reduce_axes is None: # Reached from backward() of PReLUFunctionGrad i.e. this class, to # compute higher order derivatives gW = masked else: # Reached from backward() of PReLUFunction, to compute first # derivatives gW = masked.sum(axis=self.reduce_axes) if numpy.isscalar(gW): gW = numpy.array(gW) gx = gy.copy() masked = numpy.ma.array(gx, mask=mask) masked *= W.reshape(self.extended_shape) self.retain_inputs((0, 1, 2)) return gx, gW def forward_gpu(self, inputs): x, W, gy = inputs masked = cuda.elementwise( 'T x, T cond, T gy', 'T masked', 'masked = cond >= 0 ? (T)0 : (T)(x * gy)', 'prelu_masked')(x, self.cond, gy) if self.reduce_axes is None: gW = masked.copy() else: gW = masked.sum(axis=self.reduce_axes) gx = masked # reuse buffer _fwd_kern()(gy, self.cond, W.reshape(self.extended_shape), gx) self.retain_inputs((0, 1, 2)) return gx, gW def backward(self, indexes, grad_outputs): x, W, gy = self.get_retained_inputs() ggx, ggW = grad_outputs ggW = chainer.functions.broadcast_to( chainer.functions.reshape(ggW, self.extended_shape), x.shape) ggW *= self.cond < 0 gxgy, gxW = ( PReLUFunctionGrad(self.cond, None, self.extended_shape) .apply((gy, W, ggx)) ) ret = [] if 0 in indexes: ret.append(gy * ggW) if 1 in indexes: ret.append(chainer.functions.sum(gxW, axis=self.reduce_axes)) if 2 in indexes: ret.append(x * ggW + gxgy) return ret def prelu(x, W): """Parametric ReLU function. It accepts two arguments: an input ``x`` and a weight array ``W`` and computes the output as .. math:: PReLU(x_i) = \\begin{cases} x_i & (x_i>0) \\\\ W_i * x_i & (otherwise)\\end{cases} Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. Its first axis is assumed to be the minibatch dimension. W (:class:`~chainer.Variable` or :ref:`ndarray`): Weight variable. Returns: ~chainer.Variable: Output variable .. admonition:: Example >>> x = np.arange(-3, 3, dtype=np.float32).reshape((2, 3)) >>> x array([[-3., -2., -1.], [ 0., 1., 2.]], dtype=float32) >>> W = np.array([0.01, 0.1, 1], dtype=np.float32) >>> W array([0.01, 0.1 , 1. ], dtype=float32) >>> F.prelu(x, W) variable([[-0.03, -0.2 , -1. ], [ 0. , 1. , 2. ]]) .. note:: When the PReLU function is combined with two-dimensional convolution, the elements of parameter :math:`W` are typically shared across the same filter of different pixels. In order to support such usage, this function supports the shape of parameter array that indicates leading dimensions of input arrays except the batch dimension. For example, if :math:`W` has the shape of :math:`(2, 3, 4)`, :math:`x` must have the shape of :math:`(B, 2, 3, 4, S_1, ..., S_N)` where :math:`B` is the batch size and the number of trailing :math:`S`'s :math:`N` is an arbitrary non-negative integer. .. warning:: :math:`W` is a trainable parameter in the original paper (https://arxiv.org/abs/1502.01852). To train :math:`W`, use :class:`chainer.links.PReLU` instead. .. seealso:: :class:`chainer.links.PReLU` to manage the model parameter ``W``. """ return PReLUFunction().apply((x, W))[0]
6,332
30.984848
78
py
chainer
chainer-master/chainer/functions/activation/clipped_relu.py
import numpy import chainer from chainer.backends import cuda from chainer import function_node from chainer import utils from chainer.utils import type_check import chainerx if cuda.cudnn_enabled: cudnn = cuda.cudnn _mode = cuda.cuda.cudnn.CUDNN_ACTIVATION_CLIPPED_RELU # type: ignore class ClippedReLU(function_node.FunctionNode): """Clipped Rectifier Unit function. Clipped ReLU is written as :math:`ClippedReLU(x, z) = \\min(\\max(0, x), z)`, where :math:`z(>0)` is a parameter to cap return value of ReLU. """ _use_cudnn = False def __init__(self, z): if not isinstance(z, float): raise TypeError('z must be float value') # z must be positive. assert z > 0 self.cap = z def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type = in_types[0] type_check.expect(x_type.dtype.kind == 'f') def forward_chainerx(self, inputs): x, = inputs return chainerx.clipped_relu(x, self.cap), def forward_cpu(self, inputs): self.retain_inputs((0,)) x, = inputs return utils.force_array(numpy.minimum(numpy.maximum(0, x), self.cap), x.dtype), def forward_gpu(self, inputs): self.retain_inputs((0,)) x, = inputs if chainer.should_use_cudnn('==always') and x.flags.c_contiguous: self._use_cudnn = True y = cudnn.activation_forward(x, _mode, self.cap) self.retain_outputs((0,)) else: return cuda.elementwise( 'T x, T cap', 'T y', 'y = min(max(x, (T)0), cap)', 'clipped_relu_fwd')(x, self.cap), return y, def backward(self, indexes, grad_outputs): x, = self.get_retained_inputs() if chainer.should_use_cudnn('==always') and self._use_cudnn: y = self.get_retained_outputs()[0] return ClippedReLUGrad3(x.data, y.data, self.cap).apply( grad_outputs) else: return ClippedReLUGrad2(x.data, self.cap).apply(grad_outputs) class ClippedReLUGrad2(function_node.FunctionNode): """Clipped Rectifier Unit gradient function.""" def __init__(self, x, z): self.x = x self.cap = z def check_type_forward(self, in_types): type_check._argname(in_types, ('gy',)) type_check.expect(in_types[0].dtype.kind == 'f') def forward_cpu(self, inputs): gy, = inputs x = self.x return utils.force_array( gy * (0 < x) * (x < self.cap), x.dtype), def forward_gpu(self, inputs): gy, = inputs gx = cuda.elementwise( 'T x, T gy, T z', 'T gx', 'gx = ((x > 0) & (x < z)) ? gy : (T)0', 'clipped_relu_bwd')(self.x, gy, self.cap) return gx, def backward(self, indexes, grad_outputs): return ClippedReLUGrad2(self.x, self.cap).apply(grad_outputs) class ClippedReLUGrad3(function_node.FunctionNode): """Clipped Rectifier Unit gradient function.""" def __init__(self, x, y, z): self.x = x self.y = y self.cap = z def check_type_forward(self, in_types): type_check._argname(in_types, ('gy',)) type_check.expect(in_types[0].dtype.kind == 'f') def forward_cpu(self, inputs): gy, = inputs return utils.force_array( gy * (0 < self.x) * (self.x < self.cap), self.x.dtype), def forward_gpu(self, inputs): assert chainer.should_use_cudnn('==always') return cudnn.activation_backward(self.x, self.y, inputs[0], _mode, self.cap), def backward(self, indexes, grad_outputs): return ClippedReLUGrad3(self.x, self.y, self.cap).apply(grad_outputs) def clipped_relu(x, z=20.0): """Clipped Rectifier Unit function. For a clipping value :math:`z(>0)`, it computes .. math:: \\text{ClippedReLU}(x, z) = \\min(\\max(0, x), z). Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`(s_1, s_2, ..., s_n)`-shaped float array. z (float): Clipping value. (default = 20.0) Returns: ~chainer.Variable: Output variable. A :math:`(s_1, s_2, ..., s_n)`-shaped float array. .. admonition:: Example >>> x = np.random.uniform(-100, 100, (10, 20)).astype(np.float32) >>> z = 10.0 >>> np.any(x < 0) True >>> np.any(x > z) True >>> y = F.clipped_relu(x, z=z) >>> np.any(y.array < 0) False >>> np.any(y.array > z) False """ y, = ClippedReLU(z).apply((x,)) return y def relu6(x): """Rectifier Unit function clipped at 6. It computes .. math:: \\text{ReLU6}(x) = \\min(\\max(0, x), 6). Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`(s_1, s_2, ..., s_n)`-shaped float array. Returns: ~chainer.Variable: Output variable. A :math:`(s_1, s_2, ..., s_n)`-shaped float array. .. seealso:: :func:`chainer.functions.clipped_relu` .. admonition:: Example >>> x = np.array([-20, -2, 0, 2, 4, 10, 100]).astype(np.float32) >>> x array([-20., -2., 0., 2., 4., 10., 100.], dtype=float32) >>> F.relu6(x) variable([0., 0., 0., 2., 4., 6., 6.]) """ y, = ClippedReLU(6.0).apply((x,)) return y
5,519
27.601036
78
py
chainer
chainer-master/chainer/functions/activation/sigmoid.py
import numpy import chainer from chainer.backends import cuda from chainer import function_node from chainer import utils from chainer.utils import type_check if cuda.cudnn_enabled: cudnn = cuda.cudnn _mode = cuda.libcudnn.CUDNN_ACTIVATION_SIGMOID class Sigmoid(function_node.FunctionNode): """Logistic sigmoid function.""" def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) type_check.expect(in_types[0].dtype.kind == 'f') def forward_cpu(self, inputs): x = inputs[0] half = x.dtype.type(0.5) y = utils.force_array(numpy.tanh(x * half) * half + half) self.retain_outputs((0,)) self._use_cudnn = False return y, def forward_gpu(self, inputs): x = inputs[0] if chainer.should_use_cudnn('==always') and x.flags.c_contiguous: y = cudnn.activation_forward(x, _mode) self.retain_inputs((0,)) self._use_cudnn = True else: y = cuda.elementwise( 'T x', 'T y', 'y = tanh(x * 0.5) * 0.5 + 0.5', 'sigmoid_fwd')(x) self._use_cudnn = False self.retain_outputs((0,)) return y, def backward(self, indexes, grad_outputs): if self._use_cudnn: x = self.get_retained_inputs()[0].data else: x = None y = self.get_retained_outputs()[0] gy, = grad_outputs return SigmoidGrad((x,)).apply((y, gy)) class SigmoidGrad(function_node.FunctionNode): """Logistic sigmoid gradient function.""" def __init__(self, inputs): super(SigmoidGrad, self).__init__() self.x = inputs[0] def check_type_forward(self, in_types): type_check._argname(in_types, ('y', 'gy')) type_check.expect(in_types[0].dtype.kind == 'f') type_check.expect(in_types[1].dtype.kind == 'f') def forward_cpu(self, inputs): self.retain_inputs((0, 1)) y, gy = inputs one = y.dtype.type(1) return utils.force_array(gy * y * (one - y)), def forward_gpu(self, inputs): self.retain_inputs((0, 1)) y, gy = inputs if (chainer.should_use_cudnn('==always') and gy.flags.c_contiguous and self.x is not None and self.x.flags.c_contiguous): gx = cudnn.activation_backward(self.x, y, gy, _mode) else: gx = cuda.elementwise( 'T y, T gy', 'T gx', 'gx = gy * y * (1 - y)', 'sigmoid_bwd')(y, gy) return gx, def backward(self, indexes, grad_outputs): y, gy = self.get_retained_inputs() ggx, = grad_outputs return ggx * gy * (1 - 2 * y), ggx * y * (1 - y) def sigmoid(x): """Element-wise sigmoid logistic function. .. math:: f(x)=(1 + \\exp(-x))^{-1}. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. Returns: ~chainer.Variable: Output variable. A :math:`(s_1, s_2, ..., s_N)`-shaped float array. .. admonition:: Example It maps the input values into the range of :math:`[0, 1]`. >>> x = np.arange(-2, 3, 2).astype(np.float32) >>> x array([-2., 0., 2.], dtype=float32) >>> F.sigmoid(x).array array([0.11920291, 0.5 , 0.8807971 ], dtype=float32) """ y, = Sigmoid().apply((x,)) return y
3,489
28.327731
78
py
chainer
chainer-master/chainer/functions/connection/shift.py
import numpy from chainer.backends import cuda from chainer import function_node from chainer.utils import type_check def _pair(x): if hasattr(x, '__getitem__'): return x return x, x class Shift(function_node.FunctionNode): def __init__(self, ksize=3, dilate=1): super(Shift, self).__init__() self.kh, self.kw = _pair(ksize) if self.kh % 2 != 1: raise ValueError('kh must be odd') if self.kw % 2 != 1: raise ValueError('kw must be odd') self.dy, self.dx = _pair(dilate) def check_type_forward(self, in_types): type_check._argname(in_types, ('x',)) x_type = in_types[0] type_check.expect( x_type.dtype.kind == 'f', x_type.ndim == 4, x_type.shape[1] >= self.kh * self.kw, ) def forward_cpu(self, inputs): x = inputs[0] b, c, h, w = x.shape py = self.kh // 2 * abs(self.dy) px = self.kw // 2 * abs(self.dx) x = numpy.pad(x, ((0, 0), (0, 0), (py, py), (px, px)), 'constant') n_groups = self.kh * self.kw group_size = c // n_groups ret = [] for i, group_idx in enumerate(range(n_groups)): # Make sure that center group is last if group_idx == (n_groups - 1) // 2: group_idx = n_groups - 1 elif group_idx == (n_groups - 1): group_idx = (n_groups - 1) // 2 ky = (group_idx // self.kw) - py // abs(self.dy) kx = (group_idx % self.kw) - px // abs(self.dx) hs = py + -ky * self.dy ws = px + -kx * self.dx he = hs + h we = ws + w cs = i * group_size ce = (i + 1) * group_size if i < n_groups - 1 else None ret.append(x[:, cs:ce, hs:he, ws:we]) return numpy.concatenate(ret, axis=1), def forward_gpu(self, inputs): x = inputs[0] b, c, h, w = x.shape y = cuda.cupy.empty_like(x) cuda.elementwise( 'raw T x, int32 c, int32 h, int32 w,' 'int32 kh, int32 kw,' 'int32 dy, int32 dx', 'T y', ''' int b0 = i / (c * h * w); int rest = i % (c * h * w); int c0 = rest / (h * w); rest %= h * w; int out_row = rest / w; int out_col = rest % w; int n_groups = kh * kw; int group_size = c / n_groups; int group_idx = c0 / group_size; // Make sure that center group is last if (group_idx == (n_groups - 1) / 2) { group_idx = n_groups - 1; } else if (group_idx == n_groups - 1) { group_idx = (n_groups - 1) / 2; } int ky = (group_idx / kw) - kh / 2; int kx = (group_idx % kw) - kw / 2; if (group_idx >= n_groups) { ky = 0; kx = 0; } int in_row = -ky * dy + out_row; int in_col = -kx * dx + out_col; if (in_row >= 0 && in_row < h && in_col >= 0 && in_col < w) { y = x[b0 * c * h * w + c0 * h * w + in_row * w + in_col]; } else { y = 0; } ''', 'shift_gpu')(x, c, h, w, self.kh, self.kw, self.dy, self.dx, y) return y, def backward(self, indexes, grad_outputs): return shift(grad_outputs[0], ksize=(self.kh, self.kw), dilate=(-self.dy, -self.dx)), def shift(x, ksize=3, dilate=1): """Shift function. See: `Shift: A Zero FLOP, Zero Parameter Alternative to Spatial Convolutions <https://arxiv.org/abs/1711.08141>`_ Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable of shape :math:`(n, c, h, w)`. ksize (int or pair of ints): Size of filters (a.k.a. kernels). ``ksize=k`` and ``ksize=(k, k)`` are equivalent. dilate (int or pair of ints): Dilation factor of filter applications. ``dilate=d`` and ``dilate=(d, d)`` are equivalent. Returns: ~chainer.Variable: Output variable of same shape as ``x``. """ fnode = Shift(ksize, dilate) y, = fnode.apply((x,)) return y
4,425
31.072464
77
py
chainer
chainer-master/chainer/functions/connection/deformable_convolution_2d_sampler.py
import numpy from chainer import backend from chainer.functions.array import broadcast from chainer.functions.array import concat from chainer.functions.array import pad as pad_module from chainer.functions.array import spatial_transformer_sampler from chainer.functions.math import matmul def deformable_convolution_2d_sampler(x, offset, W, b=None, stride=1, pad=0): """Two-dimensional deformable convolution function using computed offset. This is an implementation of two-dimensional deformable convolution from `Deformable Convolutional Networks <https://arxiv.org/abs/1703.06211>`_. It takes four variables: the input image ``x``, the offset image ``offset``, the filter weight ``W``, and the bias vector ``b``. Notation: here is the notation for the dimensionalities. - :math:`n` is the batch size. - :math:`c_I` and :math:`c_O` are the number of the input and output, respectively. - :math:`h` and :math:`w` are the height and width of the input image, respectively. - :math:`k_H` and :math:`k_W` are the height and width of the filters, respectively. - :math:`s_Y` and :math:`s_X` are the strides of the filter. - :math:`p_H` and :math:`p_W` are the spatial padding sizes. The output size :math:`(h_O, w_O)` is determined by the following equations: .. math:: h_O &= (h + 2p_H - k_H) / s_Y + 1,\\\\ w_O &= (w + 2p_W - k_W) / s_X + 1. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable of shape :math:`(n, c_I, h, w)`. offset (:class:`~chainer.Variable` or :ref:`ndarray`): Offset variable of shape :math:`(n, 2 \\cdot k_H \\cdot k_W, h_O, w_O)`. The first :math:`k_H \\cdot k_W` index of the second axis corresponds to the offsets in the horizontal direction. The last :math:`k_H \\cdot k_W` index of the second axis corresponds to the offsets in the vertical direction. W (:class:`~chainer.Variable` or :ref:`ndarray`): Weight variable of shape :math:`(c_O, c_I, k_H, k_W)`. b (:class:`~chainer.Variable` or :ref:`ndarray`): Bias variable of length :math:`c_O` (optional). stride (int or pair of ints): Stride of filter applications. ``stride=s`` and ``stride=(s, s)`` are equivalent. pad (int or pair of ints): Spatial padding width for input arrays. ``pad=p`` and ``pad=(p, p)`` are equivalent. Returns: ~chainer.Variable: Output variable. Deformable convolution adds 2D offsets to the regular grid sampling locations in the standard convolution. It enables free form deformation of the sampling grid. See `Jifeng Dai, Haozhi Qi, Yuwen Xiong, Yi Li, Guodong Zhang, Han Hu, Yichen Wei. Deformable Convolutional Networks <https://arxiv.org/abs/1703.06211>`_ If the bias vector is given, then it is added to all spatial locations of the output of convolution. .. seealso:: :class:`~chainer.links.DeformableConvolution2D` to manage the model parameters ``W`` and ``b``. .. admonition:: Example >>> x = np.random.uniform(0, 1, (2, 3, 4, 7)).astype(np.float32) >>> offset = np.random.uniform( ... 0, 1, (2, 2 * 3 * 3, 2, 5)).astype(np.float32) >>> W = np.random.uniform(0, 1, (4, 3, 3, 3)).astype(np.float32) >>> b = np.random.uniform(0, 1, (4,)).astype(np.float32) >>> y = F.deformable_convolution_2d_sampler(x, offset, W, b) >>> y.shape (2, 4, 2, 5) """ sy, sx = _pair(stride) ph, pw = _pair(pad) out_c, _, kh, kw = W.shape n, c, h, w = x.shape _, khkw2, out_h, out_w = offset.shape if khkw2 != 2 * kh * kw: raise ValueError( 'The shape of the offset does not match the kernel size') grid = _offset2grid(offset, kh, kw, sy, sx, ph, pw, h, w) grid = grid.reshape(n, 2, kh * kw, out_h * out_w) x_pad = pad_module.pad(x, ((0, 0), (0, 0), (ph, ph), (pw, pw)), 'constant') x_st = spatial_transformer_sampler.spatial_transformer_sampler( x_pad, grid) x_st = x_st.transpose(0, 3, 1, 2).reshape(n * out_h * out_w, c * kh * kw) W = W.transpose(1, 2, 3, 0).reshape(c * kh * kw, out_c) y = matmul.matmul(x_st, W) y = y.reshape(n, out_h, out_w, out_c).transpose(0, 3, 1, 2) if b is not None: b = broadcast.broadcast_to(b[None, :, None, None], y.shape) y += b return y def _offset2grid(offset, kh, kw, sy, sx, ph, pw, h, w): n, khkw2, out_h, out_w = offset.shape khkw = int(khkw2 / 2) xp = backend.get_array_module(offset) ys, xs = xp.meshgrid( xp.arange(0, sy * out_h, sy, dtype=numpy.float32), xp.arange(0, sx * out_w, sx, dtype=numpy.float32), indexing='ij', copy=False ) filter_offset_x = xp.tile(xp.arange(kw, dtype=numpy.float32), kh) filter_offset_y = xp.repeat(xp.arange(kh, dtype=numpy.float32), kw) x_coord = (offset[:, :khkw] + xs[None, None] + filter_offset_x[None, :, None, None]) y_coord = (offset[:, khkw:] + ys[None, None] + filter_offset_y[None, :, None, None]) # The values of this variable is clipped in range [-1, 1]. # The coordinate (-1, -1) corresponds to the upper-left # corner of the input image. x_coord = (x_coord / (w + 2 * pw - 1) - 0.5) * 2 y_coord = (y_coord / (h + 2 * ph - 1) - 0.5) * 2 # Shape of `coord` is (n, 2 * kh * kw, out_h, out_w) coord = concat.concat([x_coord, y_coord], axis=1) return coord def _pair(x): if hasattr(x, '__getitem__'): return x return x, x
5,710
37.073333
79
py
chainer
chainer-master/chainer/functions/connection/bilinear.py
import numpy import chainer from chainer import backend from chainer import function_node from chainer.utils import type_check def _as_mat(x): if x.ndim == 2: return x return x.reshape(len(x), -1) def _ij_ik_il_to_jkl(a, b, c): ab = chainer.functions.matmul(a[:, :, None], b[:, None, :]) # ijk return chainer.functions.matmul(_as_mat(ab).T, c).reshape( a.shape[1], b.shape[1], c.shape[1]) def _ij_ik_jkl_to_il(a, b, c): ab = chainer.functions.matmul(a[:, :, None], b[:, None, :]) # ijk c = c.reshape(-1, c.shape[-1]) # [jk]l return chainer.functions.matmul(_as_mat(ab), c) def _ij_il_jkl_to_ik(a, b, c): return _ij_ik_jkl_to_il(a, b, chainer.functions.swapaxes(c, 1, 2)) def _ik_il_jkl_to_ij(a, b, c): return _ij_ik_jkl_to_il(a, b, chainer.functions.rollaxis(c, 0, c.ndim)) class BilinearFunction(function_node.FunctionNode): def check_type_forward(self, in_types): n_in = type_check.eval(in_types.size()) if n_in != 3 and n_in != 6: raise type_check.InvalidType( '{0} or {1}'.format( in_types.size() == 3, in_types.size() == 6), '{0} == {1}'.format(in_types.size(), n_in)) e1_type, e2_type, W_type = in_types[:3] type_check_prod = type_check.make_variable(numpy.prod, 'prod') type_check.expect( e1_type.dtype == numpy.float32, e1_type.ndim >= 2, e2_type.dtype == numpy.float32, e2_type.ndim >= 2, e1_type.shape[0] == e2_type.shape[0], W_type.dtype == numpy.float32, W_type.ndim == 3, type_check_prod(e1_type.shape[1:]) == W_type.shape[0], type_check_prod(e2_type.shape[1:]) == W_type.shape[1], ) if n_in == 6: out_size = W_type.shape[2] V1_type, V2_type, b_type = in_types[3:] type_check.expect( V1_type.dtype == numpy.float32, V1_type.ndim == 2, V1_type.shape[0] == W_type.shape[0], V1_type.shape[1] == out_size, V2_type.dtype == numpy.float32, V2_type.ndim == 2, V2_type.shape[0] == W_type.shape[1], V2_type.shape[1] == out_size, b_type.dtype == numpy.float32, b_type.ndim == 1, b_type.shape[0] == out_size, ) def forward(self, inputs): self.retain_inputs(tuple(range(len(inputs)))) e1 = _as_mat(inputs[0]) e2 = _as_mat(inputs[1]) W = inputs[2] xp = backend.get_array_module(*inputs) # optimize: y = xp.einsum('ij,ik,jkl->il', e1, e2, W) y = xp.tensordot(xp.einsum('ij,ik->ijk', e1, e2), W, axes=2) if len(inputs) == 6: V1, V2, b = inputs[3:] y += e1.dot(V1) y += e2.dot(V2) y += b return y, def backward(self, indexes, grad_outputs): inputs = self.get_retained_inputs() e1, e2, W = inputs[:3] gy, = grad_outputs if len(inputs) == 6: V1, V2 = inputs[3], inputs[4] return BilinearFunctionGrad().apply((e1, e2, W, V1, V2, gy)) return BilinearFunctionGrad().apply((e1, e2, W, gy)) class BilinearFunctionGrad(function_node.FunctionNode): def forward(self, inputs): self.retain_inputs(tuple(range(len(inputs)))) e1 = _as_mat(inputs[0]) e2 = _as_mat(inputs[1]) W, gy = inputs[2], inputs[-1] xp = backend.get_array_module(*inputs) # optimize: gW = xp.einsum('ij,ik,il->jkl', e1, e2, gy) gW = xp.einsum('ij,ik->jki', e1, e2).dot(gy) gy_W = xp.tensordot(gy, W, axes=(1, 2)) # 'il,jkl->ijk' # optimize: ge1 = xp.einsum('ik,jkl,il->ij', e2, W, gy) ge1 = xp.einsum('ik,ijk->ij', e2, gy_W) # optimize: ge2 = xp.einsum('ij,jkl,il->ik', e1, W, gy) ge2 = xp.einsum('ij,ijk->ik', e1, gy_W) ret = ge1.reshape(inputs[0].shape), ge2.reshape(inputs[1].shape), gW if len(inputs) == 6: V1, V2 = inputs[3], inputs[4] gV1 = e1.T.dot(gy) gV2 = e2.T.dot(gy) gb = gy.sum(0) ge1 += gy.dot(V1.T) ge2 += gy.dot(V2.T) ret += gV1, gV2, gb return ret def backward(self, indexes, grad_outputs): inputs = self.get_retained_inputs() e1 = _as_mat(inputs[0]) e2 = _as_mat(inputs[1]) W, gy = inputs[2], inputs[-1] gge1 = _as_mat(grad_outputs[0]) gge2 = _as_mat(grad_outputs[1]) ggW = grad_outputs[2] dge1_de2 = _ij_il_jkl_to_ik(gge1, gy, W) dge1_dW = _ij_ik_il_to_jkl(gge1, e2, gy) dge1_dgy = _ij_ik_jkl_to_il(gge1, e2, W) dge2_de1 = _ik_il_jkl_to_ij(gge2, gy, W) dge2_dW = _ij_ik_il_to_jkl(e1, gge2, gy) dge2_dgy = _ij_ik_jkl_to_il(e1, gge2, W) dgW_de1 = _ik_il_jkl_to_ij(e2, gy, ggW) dgW_de2 = _ij_il_jkl_to_ik(e1, gy, ggW) dgW_dgy = _ij_ik_jkl_to_il(e1, e2, ggW) ge1 = dgW_de1 + dge2_de1 ge2 = dgW_de2 + dge1_de2 gW = dge1_dW + dge2_dW ggy = dgW_dgy + dge1_dgy + dge2_dgy if len(inputs) == 6: V1, V2 = inputs[3], inputs[4] ggV1, ggV2, ggb = grad_outputs[3:] gV1 = chainer.functions.matmul(gge1, gy, transa=True) gV2 = chainer.functions.matmul(gge2, gy, transa=True) ge1 += chainer.functions.matmul(gy, ggV1, transb=True) ge2 += chainer.functions.matmul(gy, ggV2, transb=True) ggy += chainer.functions.matmul(gge1, V1) ggy += chainer.functions.matmul(gge2, V2) ggy += chainer.functions.matmul(e1, ggV1) ggy += chainer.functions.matmul(e2, ggV2) ggy += chainer.functions.broadcast_to(ggb, ggy.shape) ge1 = ge1.reshape(inputs[0].shape) ge2 = ge2.reshape(inputs[1].shape) if len(inputs) == 6: return ge1, ge2, gW, gV1, gV2, ggy return ge1, ge2, gW, ggy def bilinear(e1, e2, W, V1=None, V2=None, b=None): """Applies a bilinear function based on given parameters. This is a building block of Neural Tensor Network (see the reference paper below). It takes two input variables and one or four parameters, and outputs one variable. To be precise, denote six input arrays mathematically by :math:`e^1\\in \\mathbb{R}^{I\\cdot J}`, :math:`e^2\\in \\mathbb{R}^{I\\cdot K}`, :math:`W\\in \\mathbb{R}^{J \\cdot K \\cdot L}`, :math:`V^1\\in \\mathbb{R}^{J \\cdot L}`, :math:`V^2\\in \\mathbb{R}^{K \\cdot L}`, and :math:`b\\in \\mathbb{R}^{L}`, where :math:`I` is mini-batch size. In this document, we call :math:`V^1`, :math:`V^2`, and :math:`b` linear parameters. The output of forward propagation is calculated as .. math:: y_{il} = \\sum_{jk} e^1_{ij} e^2_{ik} W_{jkl} + \\ \\sum_{j} e^1_{ij} V^1_{jl} + \\sum_{k} e^2_{ik} V^2_{kl} + b_{l}. Note that V1, V2, b are optional. If these are not given, then this function omits the last three terms in the above equation. .. note:: This function accepts an input variable ``e1`` or ``e2`` of a non-matrix array. In this case, the leading dimension is treated as the batch dimension, and the other dimensions are reduced to one dimension. .. note:: In the original paper, :math:`J` and :math:`K` must be equal and the author denotes :math:`[V^1 V^2]` (concatenation of matrices) by :math:`V`. Args: e1 (:class:`~chainer.Variable` or :ref:`ndarray`): Left input variable. e2 (:class:`~chainer.Variable` or :ref:`ndarray`): Right input variable. W (:class:`~chainer.Variable` or :ref:`ndarray`): Quadratic weight variable. V1 (:class:`~chainer.Variable` or :ref:`ndarray`): Left coefficient variable. V2 (:class:`~chainer.Variable` or :ref:`ndarray`): Right coefficient variable. b (:class:`~chainer.Variable` or :ref:`ndarray`): Bias variable. Returns: ~chainer.Variable: Output variable. See: `Reasoning With Neural Tensor Networks for Knowledge Base Completion <https://papers.nips.cc/paper/5028-reasoning-with-neural-tensor- networks-for-knowledge-base-completion>`_ [Socher+, NIPS2013]. .. seealso:: :class:`~chainer.links.Bilinear` to manage the model parameters ``W``, ``V1``, ``V2``, and ``b``. """ flags = [V1 is None, V2 is None, b is None] if any(flags): if not all(flags): raise ValueError('All coefficients and bias for bilinear() must ' 'be None, if at least one of them is None.') return BilinearFunction().apply((e1, e2, W))[0] return BilinearFunction().apply((e1, e2, W, V1, V2, b))[0]
9,015
33.412214
79
py
chainer
chainer-master/chainer/functions/connection/deconvolution_2d.py
import numpy import chainer from chainer.backends import cuda from chainer.backends import intel64 from chainer import configuration from chainer import function_node import chainer.functions from chainer.functions.connection import convolution_2d from chainer import memory_layouts from chainer.utils import argument from chainer.utils import conv from chainer.utils import type_check import chainerx if cuda.cudnn_enabled: _cudnn_version = cuda.cuda.cudnn.getVersion() # type: ignore def _pair(x): if hasattr(x, '__getitem__'): return x return x, x class Deconvolution2DFunction(function_node.FunctionNode): cover_all = None _use_ideep = False def __init__(self, stride=1, pad=0, outsize=None, **kwargs): dilate, groups = argument.parse_kwargs( kwargs, ('dilate', 1), ('groups', 1), deterministic='deterministic argument is not supported anymore. ' 'Use chainer.using_config(\'cudnn_deterministic\', value) context ' 'where value is either `True` or `False`.', requires_x_grad='requires_x_grad argument is not supported ' 'anymore. Just remove the argument. Note that whether to compute ' 'the gradient w.r.t. x is automatically decided during ' 'backpropagation.') self.sy, self.sx = _pair(stride) self.ph, self.pw = _pair(pad) self.outh, self.outw = (None, None) if outsize is None else outsize self.dy, self.dx = _pair(dilate) self.groups = groups if self.dx < 1 or self.dy < 1: raise ValueError('Dilate should be positive, but {} is ' 'supplied.'.format(dilate)) def check_type_forward(self, in_types): n_in = in_types.size() type_check.expect(2 <= n_in, n_in <= 3) x_type, w_type = in_types[:2] type_check.expect( x_type.dtype.kind == 'f', w_type.dtype.kind == 'f', x_type.ndim == 4, w_type.ndim == 4, x_type.shape[1] == w_type.shape[0] ) if self.outh is not None: lower_bound = conv.get_conv_outsize( self.outh, w_type.shape[2], self.sy, self.ph, d=self.dy) upper_bound = conv.get_conv_outsize( self.outh, w_type.shape[2], self.sy, self.ph, cover_all=True, d=self.dy) type_check.expect( lower_bound <= x_type.shape[2], x_type.shape[2] <= upper_bound) if self.outw is not None: lower_bound = conv.get_conv_outsize( self.outw, w_type.shape[3], self.sx, self.pw, d=self.dx) upper_bound = conv.get_conv_outsize( self.outw, w_type.shape[3], self.sx, self.pw, cover_all=True, d=self.dx) type_check.expect( lower_bound <= x_type.shape[3], x_type.shape[3] <= upper_bound) if type_check.eval(n_in) == 3: b_type = in_types[2] type_check.expect( b_type.dtype == x_type.dtype, b_type.ndim == 1, # Need to consider the case that group count > 1. # b_type.shape[0] == w_type.shape[1], ) def check_layout_forward(self, inputs): # TODO(niboshi): Write input layout check pass def _calc_out_size(self, x_shape, w_shape): """Calculates and stores `outh` and `outw`.""" _, _, kh, kw = w_shape _, _, in_h, in_w = x_shape # - k, m, n: shape of out_channel # - b: number of inputs # - h, w: height and width of kernels # k, m, n, b, h, w -> b, k, m, n, h, w if self.outh is None: self.outh = conv.get_deconv_outsize( in_h, kh, self.sy, self.ph, d=self.dy) if self.outh <= 0: raise RuntimeError('Height in the output must be positive.') if self.outw is None: self.outw = conv.get_deconv_outsize( in_w, kw, self.sx, self.pw, d=self.dx) if self.outw <= 0: raise RuntimeError('Width in the output must be positive.') def forward_cpu(self, inputs): if ((self.dy == 1 and self.dx == 1) and intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(inputs)): self._use_ideep = True self.retain_inputs((0, 1)) # only retain x and W if len(inputs) == 2: (x, W), b = inputs, None x_layout, w_layout = self.input_layouts else: x, W, b = inputs x_layout, w_layout, _ = self.input_layouts x_shape = memory_layouts._transpose_shape(x.shape, x_layout, None) w_shape = memory_layouts._transpose_shape(W.shape, w_layout, None) self._calc_out_size(x_shape, w_shape) if self.groups > 1: # Grouped convolution implementation return self._forward_grouped_convolution(x, W, b) elif (intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(inputs)): # iDeep implementation self._use_ideep = True return self._forward_ideep(x, W, b) else: return self._forward_cpu_core(x, W, b) def _forward_cpu_core(self, x, W, b): if self._use_ideep: return self._forward_ideep(x, W, b) gcol = numpy.tensordot(W, x, (0, 1)).astype(x.dtype, copy=False) gcol = numpy.rollaxis(gcol, 3) y = conv.col2im_cpu( gcol, self.sy, self.sx, self.ph, self.pw, self.outh, self.outw, dy=self.dy, dx=self.dx) # b, k, h, w if b is not None: y += b.reshape((1, b.size, 1, 1)) return y, def _forward_ideep(self, x, W, b): _, in_c, kh, kw = W.shape n, _, in_h, in_w = x.shape pd = (self.sy * (in_h - 1) + (kh + (kh - 1) * (self.dy - 1)) - self.outh - self.ph) pr = (self.sx * (in_w - 1) + (kw + (kw - 1) * (self.dx - 1)) - self.outw - self.pw) param = intel64.ideep.convolution2DParam( (n, in_c, self.outh, self.outw), self.dy, self.dx, self.sy, self.sx, self.ph, self.pw, pd, pr) y = intel64.ideep.convolution2D.BackwardData( intel64.ideep.array(W), intel64.ideep.array(x), param) if b is not None: y += b.reshape((1, b.size, 1, 1)) return y, def forward_gpu(self, inputs): self.retain_inputs((0, 1)) # only retain x and W if len(inputs) == 2: (x, W), b = inputs, None x_layout, w_layout = self.input_layouts else: x, W, b = inputs x_layout, w_layout, _ = self.input_layouts x_shape = memory_layouts._transpose_shape(x.shape, x_layout, None) w_shape = memory_layouts._transpose_shape(W.shape, w_layout, None) self._calc_out_size(x_shape, w_shape) self._set_cover_all(x_shape, w_shape) use_cudnn = ( chainer.should_use_cudnn('>=auto') and not self.cover_all and x.dtype == W.dtype and ((self.dy == 1 and self.dx == 1) or (_cudnn_version >= 6000 and not configuration.config.cudnn_deterministic)) and (self.groups <= 1 or _cudnn_version >= 7000) ) if use_cudnn: # cuDNN implementation return self._forward_cudnn(x, W, b, (x_layout, w_layout)) elif self.groups > 1: return self._forward_grouped_convolution(x, W, b) else: return self._forward_gpu_core(x, W, b) def _forward_gpu_core(self, x, W, b): # Implementation using col2im gcol = cuda.cupy.tensordot(W, x, (0, 1)).astype(x.dtype, copy=False) # - k, m, n: shape of out_channel # - b: number of inputs # - h, w: height and width of kernels # k, m, n, b, h, w -> b, k, m, n, h, w gcol = cuda.cupy.rollaxis(gcol, 3) y = conv.col2im_gpu( gcol, self.sy, self.sx, self.ph, self.pw, self.outh, self.outw, dy=self.dy, dx=self.dx) if b is not None: y += b.reshape(1, b.size, 1, 1) return y, def _forward_grouped_convolution(self, x, W, b): # G: group count # N: batch size # kH, kW: kernel height, kernel width # xC, xH, xW: x channels, x height, x width # yC, yH, yW: y channels, y height, y width G = self.groups N, xC, xH, xW = x.shape xCg = xC // G _, yCg, kH, kW = W.shape # _ == xC yC = yCg * G x = x.transpose(1, 0, 2, 3) # (xC, N, xH, xW) x = x.reshape(G, xCg, N * xH * xW) W = W.reshape(G, xCg, yCg * kH * kW) W = W.transpose(0, 2, 1) # (G, yCg*kH*kW, xCg) # (G, yCg*kH*kW, N*xH*xW) = (G, yCg*kH*kW, xCg) @ (G, xCg, N*xH*xW) col = convolution_2d._matmul(W, x).astype(x.dtype, copy=False) col = col.reshape(yC, kH, kW, N, xH, xW) col = col.transpose(3, 0, 1, 2, 4, 5) # (N, yC, kH, kW, xH, xW) y = conv.col2im(col, self.sy, self.sx, self.ph, self.pw, self.outh, self.outw, dy=self.dy, dx=self.dx) if b is not None: y += b.reshape(1, b.size, 1, 1) return y, def _forward_cudnn(self, x, W, b, input_layouts): x_layout, w_layout = input_layouts self.output_layouts = (x_layout,) n = len(x) _, c, _, _ = memory_layouts._transpose_shape(W.shape, w_layout, None) y_raw_shape = memory_layouts._transpose_shape( (n, c*self.groups, self.outh, self.outw), None, x_layout) y = cuda.cupy.empty(y_raw_shape, dtype=x.dtype) pad = (self.ph, self.pw) stride = (self.sy, self.sx) dilation = (self.dy, self.dx) deterministic = configuration.config.cudnn_deterministic auto_tune = configuration.config.autotune tensor_core = configuration.config.use_cudnn_tensor_core cudnn_x_layout = cuda._get_cudnn_tensor_layout_x(x_layout) cudnn_w_layout = cuda._get_cudnn_tensor_layout_w(w_layout) cuda.cudnn.convolution_backward_data( W, x, b, y, pad, stride, dilation, self.groups, deterministic=deterministic, auto_tune=auto_tune, tensor_core=tensor_core, d_layout=cudnn_x_layout, w_layout=cudnn_w_layout) return y, def forward_chainerx(self, inputs): # TODO(imanishi): Support it if self.dy != 1 or self.dx != 1: return chainer.Fallback # TODO(imanishi): Support it if self.groups != 1: return chainer.Fallback # TODO(imanishi): Support it if any(a.dtype != inputs[0].dtype for a in inputs): return chainer.Fallback # TODO(imanishi): Support it self._calc_out_size(inputs[0].shape, inputs[1].shape) self._set_cover_all(inputs[0].shape, inputs[1].shape) if self.cover_all: return chainer.Fallback stride = (self.sy, self.sx) pad = (self.ph, self.pw) outsize = None if self.outh is None else (self.outh, self.outw) return chainerx.conv_transpose( *inputs, stride=stride, pad=pad, outsize=outsize), def backward(self, indexes, grad_outputs): x, W = self.get_retained_inputs() if len(self.input_layouts) == 2: x_layout, w_layout = self.input_layouts else: x_layout, w_layout, _ = self.input_layouts gy, = grad_outputs ret = [] if 0 in indexes: if self.cover_all is None: self._set_cover_all(x.shape, W.shape) gx = chainer.functions.convolution_2d( gy, W, stride=(self.sy, self.sx), pad=(self.ph, self.pw), cover_all=self.cover_all, dilate=(self.dy, self.dx), groups=self.groups) ret.append(gx) if 1 in indexes: if self.cover_all is None: self._set_cover_all(x.shape, W.shape) gW, = convolution_2d.Convolution2DGradW( self, W.shape, W.dtype, w_layout).apply((gy, x)) ret.append(gW) if 2 in indexes: gb = chainer.functions.sum(gy, axis=(0, 2, 3)) ret.append(gb) return ret def _set_cover_all(self, x_shape, w_shape): _, _, kh, kw = w_shape _, _, in_h, in_w = x_shape self.cover_all = ( in_h != conv.get_conv_outsize(self.outh, kh, self.sy, self.ph, d=self.dy) or in_w != conv.get_conv_outsize(self.outw, kw, self.sx, self.pw, d=self.dx)) def deconvolution_2d(x, W, b=None, stride=1, pad=0, outsize=None, **kwargs): """deconvolution_2d(x, W, b=None, stride=1, pad=0, outsize=None, *, \ dilate=1, groups=1) Two dimensional deconvolution function. This is an implementation of two-dimensional deconvolution. In most of deep learning frameworks and papers, this function is called **transposed convolution**. But because of historical reasons (e.g. paper by Ziller `Deconvolutional Networks`_) and backward compatibility, this function is called **deconvolution** in Chainer. .. _Deconvolutional Networks: \ http://www.matthewzeiler.com/pubs/cvpr2010/cvpr2010.pdf It takes three variables: input image ``x``, the filter weight ``W``, and the bias vector ``b``. Notation: here is a notation for dimensionalities. - :math:`n` is the batch size. - :math:`c_I` and :math:`c_O` are the number of the input and output channels, respectively. - :math:`h_I` and :math:`w_I` are the height and width of the input image, respectively. - :math:`h_K` and :math:`w_K` are the height and width of the filters, respectively. - :math:`h_P` and :math:`w_P` are the height and width of the spatial padding size, respectively. Let :math:`(s_Y, s_X)` be the stride of filter application. Then, the output size :math:`(h_O, w_O)` is estimated by the following equations: .. math:: h_O &= s_Y (h_I - 1) + h_K - 2h_P,\\\\ w_O &= s_X (w_I - 1) + w_K - 2w_P. The output of this function can be non-deterministic when it uses cuDNN. If ``chainer.configuration.config.deterministic`` is ``True`` and cuDNN version is >= v3, it forces cuDNN to use a deterministic algorithm. Deconvolution links can use a feature of cuDNN called autotuning, which selects the most efficient CNN algorithm for images of fixed-size, can provide a significant performance boost for fixed neural nets. To enable, set `chainer.using_config('autotune', True)` Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable of shape :math:`(n, c_I, h_I, w_I)`. W (:class:`~chainer.Variable` or :ref:`ndarray`): Weight variable of shape :math:`(c_I, c_O, h_K, w_K)`. b (None or :class:`~chainer.Variable` or :ref:`ndarray`): Bias variable of length :math:`c_O` (optional). stride (:class:`int` or pair of :class:`int` s): Stride of filter applications. ``stride=s`` and ``stride=(s, s)`` are equivalent. pad (:class:`int` or pair of :class:`int` s): Spatial padding width for input arrays. ``pad=p`` and ``pad=(p, p)`` are equivalent. outsize (None or :class:`tuple` of :class:`int` s): Expected output size of deconvolutional operation. It should be pair of height and width :math:`(h_O, w_O)`. Default value is ``None`` and the outsize is estimated by input size, stride and pad. dilate (:class:`int` or pair of :class:`int` s): Dilation factor of filter applications. ``dilate=d`` and ``dilate=(d, d)`` are equivalent. groups (:class:`int`): The number of groups to use grouped deconvolution. The default is one, where grouped deconvolution is not used. Returns: ~chainer.Variable: Output variable of shape :math:`(n, c_O, h_O, w_O)`. .. seealso:: :class:`~chainer.links.Deconvolution2D` to manage the model parameters ``W`` and ``b``. .. admonition:: Example >>> n = 10 >>> c_i, c_o = 1, 3 >>> h_i, w_i = 5, 10 >>> h_k, w_k = 10, 10 >>> h_p, w_p = 5, 5 >>> x = np.random.uniform(0, 1, (n, c_i, h_i, w_i)).astype(np.float32) >>> x.shape (10, 1, 5, 10) >>> W = np.random.uniform(0, 1, (c_i, c_o, h_k, w_k)).\ astype(np.float32) >>> W.shape (1, 3, 10, 10) >>> b = np.random.uniform(0, 1, c_o).astype(np.float32) >>> b.shape (3,) >>> s_y, s_x = 5, 5 >>> y = F.deconvolution_2d(x, W, b, stride=(s_y, s_x), pad=(h_p, w_p)) >>> y.shape (10, 3, 20, 45) >>> h_o = s_y * (h_i - 1) + h_k - 2 * h_p >>> w_o = s_x * (w_i - 1) + w_k - 2 * w_p >>> y.shape == (n, c_o, h_o, w_o) True """ argument.check_unexpected_kwargs( kwargs, deterministic='deterministic argument is not ' 'supported anymore. ' 'Use chainer.using_config(\'cudnn_deterministic\', value) ' 'context where value is either `True` or `False`.') dilate, groups = argument.parse_kwargs(kwargs, ('dilate', 1), ('groups', 1)) func = Deconvolution2DFunction(stride, pad, outsize, dilate=dilate, groups=groups) if b is None: args = x, W else: args = x, W, b y, = func.apply(args) return y
18,105
36.255144
79
py
chainer
chainer-master/chainer/functions/connection/convolution_nd.py
import numpy from six import moves import chainer from chainer import backend from chainer.backends import cuda from chainer import configuration from chainer import function_node from chainer.functions.connection import convolution_2d from chainer import utils from chainer.utils import conv from chainer.utils import conv_nd from chainer.utils import type_check import chainerx class ConvolutionND(function_node.FunctionNode): def __init__(self, ndim, stride=1, pad=0, cover_all=False, dilate=1, groups=1): self.ndim = ndim self.stride = conv_nd.as_tuple(stride, ndim) self.pad = conv_nd.as_tuple(pad, ndim) self.cover_all = cover_all self.dilate = conv_nd.as_tuple(dilate, ndim) self.groups = groups def check_type_forward(self, in_types): n_in = in_types.size() type_check.expect(2 <= n_in, n_in <= 3) x_type = in_types[0] w_type = in_types[1] type_check.expect( x_type.dtype.kind == 'f', w_type.dtype.kind == 'f', x_type.ndim == self.ndim + 2, w_type.ndim == self.ndim + 2, # Need to consider the case that group count > 1. # x_type.shape[1] == w_type.shape[1], ) if type_check.eval(n_in) == 3: b_type = in_types[2] type_check.expect( b_type.dtype.kind == 'f', b_type.ndim == 1, b_type.shape[0] == w_type.shape[0], ) def forward_chainerx(self, inputs): # TODO(hvy): Support mixed precision. if any([arr.dtype != inputs[0].dtype for arr in inputs[1:]]): return chainer.Fallback # TODO(hvy): Support dilate > 1. if any(d != 1 for d in self.dilate): return chainer.Fallback # TODO(hvy): Support groups > 1. if self.groups > 1: return chainer.Fallback if inputs[0].device.backend.name == 'cuda' and ( self.cover_all or self.ndim < 2): return chainer.Fallback return chainerx.conv( *inputs, stride=self.stride, pad=self.pad, cover_all=self.cover_all), def _use_cudnn(self, x, W): if cuda._cudnn_version < 6000 and any(d != 1 for d in self.dilate): # cuDNN < 6.0 does not support dilated convolutions return False if cuda._cudnn_version < 7000 and 1 < self.groups: # cuDNN < 7.0 does not support grouped convolutions return False return ( chainer.should_use_cudnn('>=auto') and not self.cover_all and x.dtype == W.dtype and self.ndim > 1) def _forward_xp(self, x, W, b, xp): if 1 < self.groups: return self._forward_grouped_convolution_xp(x, W, b, xp) else: return self._forward_xp_core(x, W, b, xp) def _forward_grouped_convolution_xp(self, x, W, b, xp): # G: group count # N: batch size # iC: input channels # oC: output channels G = self.groups N, iC = x.shape[:2] oC = W.shape[0] k_size = W.shape[2:] iCg = iC // G oCg = oC // G dims = len(k_size) if iC % G != 0: raise TypeError('The number of groups must be ' 'a divisor of that of input channels') if oC % G != 0: raise TypeError('The number of groups must be ' 'a divisor of that of output channels') xp = backend.get_array_module(x) # (N, iC, k_size..., o_size...) x = conv_nd.im2col_nd(x, k_size, self.stride, self.pad, cover_all=self.cover_all, dilate=self.dilate) o_size = x.shape[-dims:] x = xp.rollaxis(x, 0, dims + 2) # (iC, k_size..., N, o_size...) mul_len = iCg * utils.size_of_shape(k_size) x = x.reshape(G, mul_len, N * utils.size_of_shape(o_size)) W = W.reshape(G, oCg, mul_len) # (G, oCg, N*o_size) = (G, oCg, iCg*k_size) @ (G, iCg*k_size, N*o_size) y = convolution_2d._matmul(W, x).astype(x.dtype, copy=False) y = y.reshape(oC, N, *o_size) y = xp.rollaxis(y, 1) # (N, oC, o_size...) if b is not None: y += b.reshape(1, b.size, *((1,) * dims)) return y, def _forward_xp_core(self, x, W, b, xp): ndim = self.ndim ksize = W.shape[2:] stride = self.stride pad = self.pad dilate = self.dilate # Make patch array. if xp is numpy: col = conv_nd.im2col_nd_cpu( x, ksize, stride, pad, cover_all=self.cover_all, dilate=dilate) else: col = conv_nd.im2col_nd_gpu( x, ksize, stride, pad, cover_all=self.cover_all, dilate=dilate) # Compute correlation. axes = tuple(moves.range(1, ndim + 2)) # (1, 2, ..., N+1) y = xp.tensordot(col, W, (axes, axes)).astype(x.dtype, copy=False) # Apply bias if given. if b is not None: y += b # Roll c_O before the second in (n, y_1, y_2, ..., y_N, c_O). return xp.rollaxis(y, ndim + 1, 1), def _forward_cudnn(self, x, W, b): out_c = W.shape[0] # (c_O, _, k_1, k_2, ..., k_N) ksize = W.shape[2:] n, c = x.shape[:2] # (n, c_I, d_1, d_2, ..., d_N) dims = x.shape[2:] stride = self.stride pad = self.pad dilate = self.dilate groups = self.groups # Make empty array for result. outs = tuple( conv.get_conv_outsize(d, k, s, p, cover_all=self.cover_all, d=di) for (d, k, s, p, di) in zip(dims, ksize, stride, pad, dilate)) assert all(out > 0 for out in outs), 'Output sizes should be positive.' y_shape = (n, out_c) + outs # (n, c_O, out_1, out_2, ..., out_N) y = cuda.cupy.empty(y_shape, dtype=x.dtype) auto_tune = configuration.config.autotune tensor_core = configuration.config.use_cudnn_tensor_core cuda.cudnn.convolution_forward( x, W, b, y, pad, stride, dilate, groups, auto_tune=auto_tune, tensor_core=tensor_core) return y, def forward(self, inputs): self.retain_inputs((0, 1)) # retain only x and W x, W = inputs[:2] b = inputs[2] if len(inputs) == 3 else None xp = backend.get_array_module(*inputs) if xp is numpy: return self._forward_xp(x, W, b, numpy) elif not self._use_cudnn(x, W): return self._forward_xp(x, W, b, cuda.cupy) else: return self._forward_cudnn(x, W, b) def backward(self, indexes, grad_outputs): x, W = self.get_retained_inputs() gy, = grad_outputs ret = [] if 0 in indexes: x_shape = x.shape[2:] gx = chainer.functions.deconvolution_nd( gy, W, stride=self.stride, pad=self.pad, outsize=x_shape, dilate=self.dilate, groups=self.groups) ret.append(gx) if 1 in indexes: gW, = ConvolutionNDGradW(self).apply((x, gy)) ret.append(gW) if 2 in indexes: axis = (0,) + tuple(moves.range(2, gy.ndim)) gb = chainer.functions.sum(gy, axis=axis) if gb.dtype != self.inputs[2].dtype: gb = chainer.functions.cast(gb, self.inputs[2].dtype) ret.append(gb) return ret class ConvolutionNDGradW(function_node.FunctionNode): def __init__(self, convnd): W_node = convnd.inputs[1] self.ndim = convnd.ndim self.ksize = W_node.shape[2:] self.stride = convnd.stride self.pad = convnd.pad self.cover_all = convnd.cover_all self.dilate = convnd.dilate self.groups = convnd.groups self.W_dtype = W_node.dtype def _use_cudnn(self, x, gy): if cuda._cudnn_version < 6000 and any(d != 1 for d in self.dilate): # cuDNN < 6.0 does not support dilated convolutions return False if cuda._cudnn_version < 7000 and 1 < self.groups: # cuDNN < 7.0 does not support grouped convolutions return False return ( chainer.should_use_cudnn('>=auto') and not self.cover_all and x.dtype == self.W_dtype and gy.dtype == self.W_dtype and self.ndim > 1) def forward(self, inputs): self.retain_inputs((0, 1)) x, gy = inputs xp = backend.get_array_module(*inputs) if xp is numpy: return self._forward_xp(x, gy, numpy) elif not self._use_cudnn(x, gy): return self._forward_xp(x, gy, cuda.cupy) else: return self._forward_cudnn(x, gy) def _forward_xp(self, x, gy, xp): if 1 < self.groups: return self._forward_grouped_convolution_xp(x, gy, xp) else: return self._forward_xp_core(x, gy, xp) def _forward_grouped_convolution_xp(self, x, gy, xp): G = self.groups N, iC = x.shape[:2] oC = gy.shape[1] o_size = gy.shape[2:] o_size_prod = utils.size_of_shape(o_size) k_size = self.ksize dims = len(o_size) iCg = iC // G oCg = oC // G # Do not check iCg and oCg because this class is rarely used alone # (N, iC, k_size..., o_size...) x = conv_nd.im2col_nd(x, k_size, self.stride, self.pad, cover_all=self.cover_all, dilate=self.dilate) x = xp.rollaxis(x, 0, dims + 2) # (iC, k_size..., N, o_size...) mul_len = iCg * utils.size_of_shape(k_size) x = x.reshape(G, mul_len, N * o_size_prod) x = x.transpose(0, 2, 1) # (G, N*o_size, iCg*k_size) gy = xp.rollaxis(gy, 1) # (oC, N, o_size...) gy = gy.reshape(G, oCg, N * o_size_prod) # (G, oCg, iCg*k_size) = (G, oCg, N*o_size) @ (G, N*o_size, iCg*k_size) gW = convolution_2d._matmul(gy, x).astype(self.W_dtype, copy=False) gW = gW.reshape(oC, iCg, *k_size) return gW, def _forward_xp_core(self, x, gy, xp): # Compute filter weight gradient. # (n, _, out_1, out_2, ..., out_N) out_axes = (0,) + tuple(moves.range(2, self.ndim + 2)) # (n, _, _, ..., _, out_1, out_2, ..., out_N) col_axes = (0,) + tuple(moves.range(self.ndim + 2, self.ndim * 2 + 2)) # NumPy raises an error when the array is not contiguous. # See: https://github.com/chainer/chainer/issues/2744 # TODO(niboshi): Remove this code when NumPy is fixed. if (xp is numpy and not (gy.flags.c_contiguous or gy.flags.f_contiguous) and 1 in gy.shape): gy = numpy.ascontiguousarray(gy) if xp is numpy: col = conv_nd.im2col_nd_cpu( x, self.ksize, self.stride, self.pad, cover_all=self.cover_all, dilate=self.dilate) else: col = conv_nd.im2col_nd_gpu( x, self.ksize, self.stride, self.pad, cover_all=self.cover_all, dilate=self.dilate) gW = xp.tensordot(gy, col, (out_axes, col_axes)).astype( self.W_dtype, copy=False) return gW, def _forward_cudnn(self, x, gy): # Make empty arrays for result. out_c = gy.shape[1] in_c = x.shape[1] // self.groups gW = cuda.cupy.empty( (out_c, in_c) + self.ksize, dtype=self.W_dtype) # Compute pad = self.pad stride = self.stride dilate = self.dilate groups = self.groups deterministic = configuration.config.cudnn_deterministic auto_tune = configuration.config.autotune tensor_core = configuration.config.use_cudnn_tensor_core cuda.cudnn.convolution_backward_filter( x, gy, gW, pad, stride, dilate, groups, deterministic=deterministic, auto_tune=auto_tune, tensor_core=tensor_core) return gW, def backward(self, indexes, grad_outputs): x, gy = self.get_retained_inputs() ggW, = grad_outputs ret = [] if 0 in indexes: x_shape = x.shape[2:] gx = chainer.functions.deconvolution_nd( gy, ggW, stride=self.stride, pad=self.pad, outsize=x_shape, groups=self.groups, dilate=self.dilate) ret.append(gx) if 1 in indexes: ggy = convolution_nd( x, ggW, stride=self.stride, pad=self.pad, cover_all=self.cover_all, groups=self.groups, dilate=self.dilate) ret.append(ggy) return ret def convolution_nd(x, W, b=None, stride=1, pad=0, cover_all=False, dilate=1, groups=1): """N-dimensional convolution function. This is an implementation of N-dimensional convolution which is generalized two-dimensional convolution in ConvNets. It takes three variables: the input ``x``, the filter weight ``W`` and the bias vector ``b``. Notation: here is a notation for dimensionalities. - :math:`N` is the number of spatial dimensions. - :math:`n` is the batch size. - :math:`c_I` and :math:`c_O` are the number of the input and output channels, respectively. - :math:`d_1, d_2, ..., d_N` are the size of each axis of the input's spatial dimensions, respectively. - :math:`k_1, k_2, ..., k_N` are the size of each axis of the filters, respectively. - :math:`l_1, l_2, ..., l_N` are the size of each axis of the output's spatial dimensions, respectively. - :math:`p_1, p_2, ..., p_N` are the size of each axis of the spatial padding size, respectively. Then the ``convolution_nd`` function computes correlations between filters and patches of size :math:`(k_1, k_2, ..., k_N)` in ``x``. Note that correlation here is equivalent to the inner product between expanded tensors. Patches are extracted at positions shifted by multiples of ``stride`` from the first position ``(-p_1, -p_2, ..., -p_N)`` for each spatial axis. Let :math:`(s_1, s_2, ..., s_N)` be the stride of filter application. Then, the output size :math:`(l_1, l_2, ..., l_N)` is determined by the following equations: .. math:: l_n = (d_n + 2p_n - k_n) / s_n + 1 \\ \\ (n = 1, ..., N) If ``cover_all`` option is ``True``, the filter will cover the all spatial locations. So, if the last stride of filter does not cover the end of spatial locations, an additional stride will be applied to the end part of spatial locations. In this case, the output size is determined by the following equations: .. math:: l_n = (d_n + 2p_n - k_n + s_n - 1) / s_n + 1 \\ \\ (n = 1, ..., N) Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable of shape :math:`(n, c_I, d_1, d_2, ..., d_N)`. W (:class:`~chainer.Variable` or :ref:`ndarray`): Weight variable of shape :math:`(c_O, c_I, k_1, k_2, ..., k_N)`. b (None or :class:`~chainer.Variable` or :ref:`ndarray`): One-dimensional bias variable with length :math:`c_O` (optional). stride (:class:`int` or :class:`tuple` of :class:`int` s): Stride of filter applications :math:`(s_1, s_2, ..., s_N)`. ``stride=s`` is equivalent to ``(s, s, ..., s)``. pad (:class:`int` or :class:`tuple` of :class:`int` s): Spatial padding width for input arrays :math:`(p_1, p_2, ..., p_N)`. ``pad=p`` is equivalent to ``(p, p, ..., p)``. cover_all (bool): If ``True``, all spatial locations are convoluted into some output pixels. It may make the output size larger. `cover_all` needs to be ``False`` if you want to use cuDNN. dilate (:class:`int` or :class:`tuple` of :class:`int` s): Dilation factor of filter applications. ``dilate=d`` and ``dilate=(d, d, ..., d)`` are equivalent. groups (:class:`int`): The number of groups to use grouped convolution. The default is one, where grouped convolution is not used. Returns: ~chainer.Variable: Output variable of shape :math:`(n, c_O, l_1, l_2, ..., l_N)`. .. note:: This function uses cuDNN implementation for its forward and backward computation if ALL of the following conditions are satisfied: - ``cuda.cudnn_enabled`` is ``True`` - ``chainer.config.use_cudnn`` is ``'always'`` or ``'auto'`` - The number of spatial dimensions is more than one. - ``cover_all`` is ``False`` - The input's ``dtype`` is equal to the filter weight's. - The ``dtype`` is FP16, FP32 or FP64. (FP16 is only available when cuDNN version :math:`\\geq` v3.) Convolution links can use a feature of cuDNN called autotuning, which selects the most efficient CNN algorithm for images of fixed-size, can provide a significant performance boost for fixed neural nets. To enable, set `chainer.using_config('autotune', True)` .. seealso:: :class:`~chainer.links.ConvolutionND` to manage the model parameters ``W`` and ``b``. .. seealso:: :func:`convolution_2d` .. admonition:: Example >>> n = 10 >>> c_i, c_o = 3, 1 >>> d1, d2, d3 = 30, 40, 50 >>> k1, k2, k3 = 10, 10, 10 >>> p1, p2, p3 = 5, 5, 5 >>> x = np.random.uniform(0, 1, (n, c_i, d1, d2, d3)).\ astype(np.float32) >>> x.shape (10, 3, 30, 40, 50) >>> W = np.random.uniform(0, 1, (c_o, c_i, k1, k2, k3)).\ astype(np.float32) >>> W.shape (1, 3, 10, 10, 10) >>> b = np.random.uniform(0, 1, (c_o)).astype(np.float32) >>> b.shape (1,) >>> s1, s2, s3 = 2, 4, 6 >>> y = F.convolution_nd(x, W, b, stride=(s1, s2, s3),\ pad=(p1, p2, p3)) >>> y.shape (10, 1, 16, 11, 9) >>> l1 = int((d1 + 2 * p1 - k1) / s1 + 1) >>> l2 = int((d2 + 2 * p2 - k2) / s2 + 1) >>> l3 = int((d3 + 2 * p3 - k3) / s3 + 1) >>> y.shape == (n, c_o, l1, l2, l3) True >>> y = F.convolution_nd(x, W, b, stride=(s1, s2, s3),\ pad=(p1, p2, p3), cover_all=True) >>> y.shape == (n, c_o, l1, l2, l3 + 1) True """ ndim = len(x.shape[2:]) fnode = ConvolutionND( ndim, stride, pad, cover_all, dilate=dilate, groups=groups) args = (x, W) if b is None else (x, W, b) y, = fnode.apply(args) return y def convolution_1d(x, W, b=None, stride=1, pad=0, cover_all=False, dilate=1, groups=1): """1-dimensional convolution function. .. note:: This function calls :func:`~chainer.functions.convolution_nd` internally, so see the details of the behavior in the documentation of :func:`~chainer.functions.convolution_nd`. """ if len(x.shape[2:]) != 1: raise ValueError( 'The number of dimensions under channel dimension of the input ' '\'x\' should be 1. But the actual ndim was {}.'.format( len(x.shape[2:]))) return convolution_nd(x, W, b, stride, pad, cover_all, dilate, groups) def convolution_3d(x, W, b=None, stride=1, pad=0, cover_all=False, dilate=1, groups=1): """3-dimensional convolution function. .. note:: This function calls :func:`~chainer.functions.convolution_nd` internally, so see the details of the behavior in the documentation of :func:`~chainer.functions.convolution_nd`. """ if len(x.shape[2:]) != 3: raise ValueError( 'The number of dimensions under channel dimension of the input ' '\'x\' should be 3. But the actual ndim was {}.'.format( len(x.shape[2:]))) return convolution_nd(x, W, b, stride, pad, cover_all, dilate, groups)
20,150
36.247689
79
py
chainer
chainer-master/chainer/functions/connection/linear.py
import numpy from chainer import backend from chainer.backends import intel64 from chainer import function_node import chainer.functions from chainer.graph_optimizations import static_code from chainer import utils from chainer.utils import type_check import chainerx class LinearFunction(function_node.FunctionNode): _config_use_ideep = None _supports_static_optimizations = True def check_type_forward(self, in_types): n_in = in_types.size() type_check.expect(2 <= n_in, n_in <= 3) x_type, w_type = in_types[:2] type_check._argname((x_type, w_type), ('x', 'W')) type_check.expect( x_type.dtype.kind == 'f', w_type.dtype.kind == 'f', x_type.ndim == 2, w_type.ndim == 2, x_type.shape[1] == w_type.shape[1], ) if type_check.eval(n_in) == 3: b_type = in_types[2] type_check._argname((b_type,), ('b',)) type_check.expect( b_type.dtype == x_type.dtype, b_type.ndim == 1, b_type.shape[0] == w_type.shape[0], ) @static_code def static_linear_no_bias(self, xp, optimized, inputs, outputs): x, W = inputs y = outputs[0] # NumPy raises an error when the array is not contiguous. # See: https://github.com/chainer/chainer/issues/2744 # TODO(niboshi): Remove this code when NumPy is fixed. if (isinstance(x, numpy.ndarray) and not (x.flags.c_contiguous or x.flags.f_contiguous) and 1 in x.shape): x = numpy.ascontiguousarray(x) if optimized: # Note: We can only call this function when both x and W # have the same dtype. Otherwise, the output type (for y) # may not be as expected (i.e., not the same dtype as x). xp.dot(x, W.T, out=y) else: y[:] = x.dot(W.T).astype(x.dtype, copy=False) @static_code def static_add_bias(self, inputs, outputs): bias = inputs[0] y = outputs[0] y += bias def forward_chainerx(self, inputs): # TODO(niboshi): Support dtype casting in ChainerX if inputs[0].dtype != inputs[1].dtype: return chainer.Fallback # Generic implementation if len(inputs) == 3: x, W, b = inputs if x.dtype != b.dtype: return chainer.Fallback return chainerx.linear(x, W, b), else: x, W = inputs return chainerx.linear(x, W), def forward(self, inputs): self._config_use_ideep = chainer.config.use_ideep if (intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(inputs)): # iDeep implementation return self._forward_ideep(inputs) # Generic implementation if len(inputs) == 3: x, W, b = inputs else: (x, W), b = inputs, None # NumPy raises an error when the array is not contiguous. # See: https://github.com/chainer/chainer/issues/2744 # TODO(niboshi): Remove this code when NumPy is fixed. if (isinstance(x, numpy.ndarray) and not (x.flags.c_contiguous or x.flags.f_contiguous) and 1 in x.shape): x = numpy.ascontiguousarray(x) # In order to be compatible with the "static graph" feature, it is # required that all output arrays of this forward # function be allocated explicitly: xp = backend.get_array_module(x) y = xp.empty((x.shape[0], W.shape[0]), dtype=x.dtype) # This is required because all of the "static_*()" functions # use the convention that any output arrays are supplied # as input arguments to the function. That is because it is # not allowed for a "static_*()" function to return anything # other than `None`. The reason is to prevent dynamic allocation # of output arrays during execution of the static schedule # because it would break the model. self.static_linear_no_bias(xp, x.dtype == W.dtype, inputs=[x, W], outputs=[y]) if len(inputs) == 3: self.static_add_bias(inputs=[b], outputs=[y]) self.retain_inputs((0, 1)) # b is not retained return y, def _forward_ideep(self, inputs): if len(inputs) == 3: x, W, b = inputs else: (x, W), b = inputs, None y = intel64.ideep.linear.Forward( intel64.ideep.array(x), intel64.ideep.array(W), intel64.ideep.array(b) if b is not None else None) self.retain_inputs((0, 1)) return y, def backward(self, indexes, grad_outputs): x, W = self.get_retained_inputs() gy, = grad_outputs ret = [] with chainer.using_config('use_ideep', self._config_use_ideep): if 0 in indexes: gx, = LinearGradData().apply((W, gy)) ret.append(chainer.functions.cast(gx, x.dtype)) if 1 in indexes: gW, = LinearGradWeight(W.dtype).apply((x, gy)) ret.append(chainer.functions.cast(gW, W.dtype)) if 2 in indexes: gb = chainer.functions.sum(gy, axis=0) ret.append(gb) return ret class LinearGradData(function_node.FunctionNode): _config_use_ideep = None def forward(self, inputs): self._config_use_ideep = chainer.config.use_ideep if (intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(inputs)): # iDeep implementation return self._forward_ideep(inputs) # Generic implementation self.retain_inputs((0, 1)) W, gy = inputs if (isinstance(gy, numpy.ndarray) and not (gy.flags.c_contiguous or gy.flags.f_contiguous) and 1 in gy.shape): gy = numpy.ascontiguousarray(gy) gx = gy.dot(W).astype(gy.dtype, copy=False) return gx, def _forward_ideep(self, inputs): self.retain_inputs((0, 1)) W, gy = inputs gx = intel64.ideep.linear.BackwardData( intel64.ideep.array(W), intel64.ideep.array(gy)) return gx, def backward(self, indexes, grad_outputs): W, gy = self.get_retained_inputs() ggx, = grad_outputs ret = [] with chainer.using_config('use_ideep', self._config_use_ideep): if 0 in indexes: gw, = LinearGradWeight(W.dtype).apply((ggx, gy)) ret.append(chainer.functions.cast(gw, W.dtype)) if 1 in indexes: ggy = linear(ggx, W) ret.append(chainer.functions.cast(ggy, gy.dtype)) return ret class LinearGradWeight(function_node.FunctionNode): _config_use_ideep = None def __init__(self, w_dtype): self._w_dtype = w_dtype def forward(self, inputs): self._config_use_ideep = chainer.config.use_ideep if (intel64.should_use_ideep('>=auto') and self._w_dtype == numpy.float32 and intel64.inputs_all_ready(inputs)): # iDeep implementation return self._forward_ideep(inputs) # Generic implementation self.retain_inputs((0, 1)) x, gy = inputs if (isinstance(gy, numpy.ndarray) and not (gy.flags.c_contiguous or gy.flags.f_contiguous) and 1 in gy.shape): gy = numpy.ascontiguousarray(gy) gW = gy.T.dot(x).astype(self._w_dtype, copy=False) return gW, def _forward_ideep(self, inputs): self.retain_inputs((0, 1)) x, gy = inputs gW = intel64.ideep.linear.BackwardWeights( intel64.ideep.array(x), intel64.ideep.array(gy)) return gW, def backward(self, indexes, grad_outputs): x, gy = self.get_retained_inputs() ggW, = grad_outputs ret = [] with chainer.using_config('use_ideep', self._config_use_ideep): if 0 in indexes: gx, = LinearGradData().apply((ggW, gy)) ret.append(chainer.functions.cast(gx, x.dtype)) if 1 in indexes: ggy = linear(x, ggW) ret.append(chainer.functions.cast(ggy, gy.dtype)) return ret def linear(x, W, b=None, n_batch_axes=1): """Linear function, or affine transformation. It accepts two or three arguments: an input minibatch ``x``, a weight matrix ``W``, and optionally a bias vector ``b``. It computes .. math:: y_i = W x_i + b. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable, which is a :math:`(s_1, s_2, ..., s_n)`-shaped float array. Its first ``n_batch_axes`` dimensions are handled as *minibatch dimensions*. The other dimensions are handled as concatenated one dimension whose size must be :math:`(s_{\\rm n\\_batch\\_axes} * ... * s_n = N)`. W (:class:`~chainer.Variable` or :ref:`ndarray`): Weight variable of shape :math:`(M, N)`, where :math:`(N = s_{\\rm n\\_batch\\_axes} * ... * s_n)`. b (:class:`~chainer.Variable` or :ref:`ndarray`): Bias variable (optional) of shape :math:`(M,)`. n_batch_axes (int): The number of batch axes. The default is 1. The input variable is reshaped into (:math:`{\\rm n\\_batch\\_axes} + 1`)-dimensional tensor. This should be greater than 0. Returns: ~chainer.Variable: Output variable. A float array with shape of :math:`(s_1, ..., s_{\\rm n\\_batch\\_axes}, M)`. .. seealso:: :class:`~chainer.links.Linear` to manage the model parameters ``W`` and ``b``. .. admonition:: Example >>> x = np.random.uniform(0, 1, (3, 4)).astype(np.float32) >>> W = np.random.uniform(0, 1, (5, 4)).astype(np.float32) >>> b = np.random.uniform(0, 1, (5,)).astype(np.float32) >>> y = F.linear(x, W, b) >>> y.shape (3, 5) """ if n_batch_axes <= 0: raise ValueError('n_batch_axes should be greater than 0.') if n_batch_axes > 1: batch_shape = x.shape[:n_batch_axes] batch_size = utils.size_of_shape(batch_shape) x = x.reshape(batch_size, -1) elif x.ndim > 2: x = x.reshape(x.shape[0], -1) if b is None: args = x, W else: args = x, W, b y, = LinearFunction().apply(args) if n_batch_axes > 1: y = y.reshape(batch_shape + (-1,)) return y
10,790
33.586538
75
py
chainer
chainer-master/chainer/functions/connection/deconvolution_nd.py
import numpy from six import moves import chainer from chainer import backend from chainer.backends import cuda from chainer import configuration from chainer import function_node from chainer.functions.connection import convolution_2d from chainer.functions.connection import convolution_nd from chainer import utils from chainer.utils import conv from chainer.utils import conv_nd from chainer.utils import type_check import chainerx class DeconvolutionND(function_node.FunctionNode): cover_all = None def __init__(self, ndim, stride=1, pad=0, outsize=None, dilate=1, groups=1): self.ndim = ndim self.stride = conv_nd.as_tuple(stride, ndim) self.pad = conv_nd.as_tuple(pad, ndim) if outsize is not None: assert len(outsize) == ndim self.outs = outsize self.dilate = conv_nd.as_tuple(dilate, ndim) self.groups = groups def check_type_forward(self, in_types): n_in = in_types.size() type_check.expect(2 <= n_in, n_in <= 3) x_type, w_type = in_types[:2] type_check.expect( x_type.dtype.kind == 'f', w_type.dtype.kind == 'f', x_type.ndim == self.ndim + 2, w_type.ndim == self.ndim + 2, x_type.shape[1] == w_type.shape[0] ) if self.outs is not None: for i, (out, s, p, di) in enumerate(zip( self.outs, self.stride, self.pad, self.dilate)): lower_bound = conv.get_conv_outsize( out, w_type.shape[i + 2], s, p, d=di) upper_bound = conv.get_conv_outsize( out, w_type.shape[i + 2], s, p, cover_all=True, d=di) type_check.expect( lower_bound <= x_type.shape[i + 2], x_type.shape[i + 2] <= upper_bound) if type_check.eval(n_in) == 3: b_type = in_types[2] type_check.expect( b_type.dtype.kind == 'f', b_type.ndim == 1, # Need to consider the case that group count > 1. # b_type.shape[0] == w_type.shape[1] ) def _use_cudnn(self, x, W, b): if ((cuda._cudnn_version < 6000 or configuration.config.cudnn_deterministic) and any(d != 1 for d in self.dilate)): # cuDNN < 6.0 and deterministic algorithms # does not support dilated convolutions return False if cuda._cudnn_version < 7000 and 1 < self.groups: # cuDNN < 7.0 does not support grouped convolutions return False return ( chainer.should_use_cudnn('>=auto') and not self.cover_all and self.ndim > 1 and x.dtype == W.dtype and (b is None or x.dtype == b.dtype)) def _forward_xp(self, x, W, b, xp): if 1 < self.groups: return self._forward_grouped_convolution_xp(x, W, b, xp) else: return self._forward_xp_core(x, W, b, xp) def _forward_grouped_convolution_xp(self, x, W, b, xp): # G: group count # N: batch size # xC: input channels # yC: output channels G = self.groups N, xC = x.shape[:2] x_size = x.shape[2:] yCg = W.shape[1] yC = yCg * G xCg = xC // G k_size = W.shape[2:] dims = len(k_size) if xC % G != 0: raise TypeError('The number of groups must be ' 'a divisor of that of input channels') x = xp.rollaxis(x, 1) # (xC, N, x_size...) x = x.reshape(G, xCg, N * utils.size_of_shape(x_size)) W = W.reshape(G, xCg, yCg * utils.size_of_shape(k_size)) W = W.transpose(0, 2, 1) # (G, yCg*k_size, xCg) # (G, yCg*k_size, N*x_size) = (G, yCg*k_size, xCg) @ (G, xCg, N*x_size) col = convolution_2d._matmul(W, x).astype(x.dtype, copy=False) col = col.reshape((yC,) + k_size + (N,) + x_size) col = xp.rollaxis(col, dims + 1) # (N, yC, k_size..., x_size...) y = conv_nd.col2im_nd(col, self.stride, self.pad, self.outs, dilate=self.dilate) if b is not None: y += b.reshape(1, yC, *((1,) * dims)) return y, def _forward_xp_core(self, x, W, b, xp): ndim = self.ndim stride = self.stride pad = self.pad dilate = self.dilate # gcol: C_O, k_1, ..., k_N, n, d_1, ..., d_N gcol = xp.tensordot(W, x, (0, 1)).astype(x.dtype, copy=False) # Roll n, which is batch size, before the first. gcol = xp.rollaxis(gcol, ndim + 1) # y: n, C_O, d_1, d_2, ..., d_N if xp is numpy: y = conv_nd.col2im_nd_cpu( gcol, stride, pad, self.outs, dilate=dilate) else: y = conv_nd.col2im_nd_gpu( gcol, stride, pad, self.outs, dilate=dilate) if b is not None: b_shape = (1, -1) + (1,) * ndim y += b.reshape(b_shape) return y, def _forward_cudnn(self, x, W, b): c = W.shape[1] * self.groups n, in_c = x.shape[:2] # x: n, C_I, d_1, d_2, ..., d_N # Make empty array for output. y_shape = (n, c) + self.outs # (n, c_O, out_1, out_2, ..., out_N) y = cuda.cupy.empty(y_shape, dtype=x.dtype) pad = self.pad stride = self.stride dilate = self.dilate groups = self.groups deterministic = configuration.config.cudnn_deterministic auto_tune = configuration.config.autotune tensor_core = configuration.config.use_cudnn_tensor_core cuda.cudnn.convolution_backward_data( W, x, b, y, pad, stride, dilate, groups, deterministic=deterministic, auto_tune=auto_tune, tensor_core=tensor_core) return y, def forward_chainerx(self, inputs): # TODO(imanishi): Support it if any(d != 1 for d in self.dilate): return chainer.Fallback # TODO(imanishi): Support it if self.groups != 1: return chainer.Fallback # TODO(imanishi): Support it if any(a.dtype != inputs[0].dtype for a in inputs): return chainer.Fallback # TODO(imanishi): Supporft it if inputs[0].device.backend.name == 'cuda' and self.ndim < 2: return chainer.Fallback stride = self.stride pad = self.pad return chainerx.conv_transpose(*inputs, stride=stride, pad=pad), def forward(self, inputs): self.retain_inputs((0, 1)) # only retain x and W x, W = inputs[:2] b = inputs[2] if len(inputs) == 3 else None if self.outs is None: dims = x.shape[2:] ksize = W.shape[2:] self.outs = tuple( conv.get_deconv_outsize(d, k, s, p, d=di) for d, k, s, p, di in zip(dims, ksize, self.stride, self.pad, self.dilate)) assert all(out > 0 for out in self.outs), \ 'Output sizes should be positive.' self._set_cover_all(x, W) xp = backend.get_array_module(*inputs) if xp is numpy: return self._forward_xp(x, W, b, numpy) elif not self._use_cudnn(x, W, b): return self._forward_xp(x, W, b, cuda.cupy) else: return self._forward_cudnn(x, W, b) def backward(self, indexes, grad_outputs): x, W = self.get_retained_inputs() gy, = grad_outputs ret = [] if 0 in indexes: gx = chainer.functions.convolution_nd( gy, W, stride=self.stride, pad=self.pad, cover_all=self.cover_all, dilate=self.dilate, groups=self.groups) ret.append(gx) if 1 in indexes: gW, = convolution_nd.ConvolutionNDGradW(self).apply((gy, x)) ret.append(gW) if 2 in indexes: axis = (0,) + tuple(moves.range(2, gy.ndim)) gb = chainer.functions.sum(gy, axis=axis) if gb.dtype != self.inputs[2].dtype: gb = chainer.functions.cast(gb, self.inputs[2].dtype) ret.append(gb) return ret def _set_cover_all(self, x, W): x_shape = x.shape[2:] k_shape = W.shape[2:] self.cover_all = any( ix != conv.get_conv_outsize(oy, k, s, p, d=di) for (ix, oy, k, s, p, di) in zip(x_shape, self.outs, k_shape, self.stride, self.pad, self.dilate)) def deconvolution_nd(x, W, b=None, stride=1, pad=0, outsize=None, dilate=1, groups=1): """N-dimensional deconvolution function. This is an implementation of N-dimensional deconvolution which generalizes two-dimensional one. In most of deep learning frameworks and papers, this function is called **transposed convolution**. But because of historical reasons (e.g. paper by Ziller `Deconvolutional Networks`_) and backward compatibility, this function is called **deconvolution** in Chainer. .. _Deconvolutional Networks: \ http://www.matthewzeiler.com/pubs/cvpr2010/cvpr2010.pdf It takes three variables: the input ``x``, the filter weight ``W``, and the bias vector ``b``. Notation: here is a notation for dimensionalities. - :math:`N` is the number of spatial dimensions. - :math:`n` is the batch size. - :math:`c_I` and :math:`c_O` are the number of the input and output channels, respectively. - :math:`d_1, d_2, ..., d_N` are the size of each axis of the input's spatial dimensions, respectively. - :math:`k_1, k_2, ..., k_N` are the size of each axis of the filters, respectively. - :math:`p_1, p_2, ..., p_N` are the size of each axis of the spatial padding size, respectively. - :math:`s_1, s_2, ..., s_N` are the stride of each axis of filter application, respectively. If ``outsize`` option is ``None``, the output size :math:`(l_1, l_2, ..., l_N)` is determined by the following equations with the items in the above list: .. math:: l_n = s_n (d_n - 1) + k_n - 2 p_n \\ \\ (n = 1, ..., N) If ``outsize`` option is given, the output size is determined by ``outsize``. In this case, the ``outsize`` :math:`(l_1, l_2, ..., l_N)` must satisfy the following equations: .. math:: d_n = \\lfloor (l_n + 2p_n - k_n) / s_n \\rfloor + 1 \\ \\ \ (n = 1, ..., N) Deconvolution links can use a feature of cuDNN called autotuning, which selects the most efficient CNN algorithm for images of fixed-size, can provide a significant performance boost for fixed neural nets. To enable, set `chainer.using_config('autotune', True)` Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable of shape :math:`(n, c_I, d_1, d_2, ..., d_N)`. W (:class:`~chainer.Variable` or :ref:`ndarray`): Weight variable of shape :math:`(c_I, c_O, k_1, k_2, ..., k_N)`. b (None or :class:`~chainer.Variable` or :ref:`ndarray`): One-dimensional bias variable with length :math:`c_O` (optional). stride (:class:`int` or :class:`tuple` of :class:`int` s): Stride of filter applications :math:`(s_1, s_2, ..., s_N)`. ``stride=s`` is equivalent to ``(s, s, ..., s)``. pad (:class:`int` or :class:`tuple` of :class:`int` s): Spatial padding width for input arrays :math:`(p_1, p_2, ..., p_N)`. ``pad=p`` is equivalent to ``(p, p, ..., p)``. outsize (None or :class:`tuple` of :class:`int` s): Expected output size of deconvolutional operation. It should be a tuple of ints :math:`(l_1, l_2, ..., l_N)`. Default value is ``None`` and the outsize is estimated by input size, stride and pad. dilate (:class:`int` or :class:`tuple` of :class:`int` s): Dilation factor of filter applications. ``dilate=d`` and ``dilate=(d, d, ..., d)`` are equivalent. groups (:class:`int`): The number of groups to use grouped convolution. The default is one, where grouped convolution is not used. Returns: ~chainer.Variable: Output variable of shape :math:`(n, c_O, l_1, l_2, ..., l_N)`. .. seealso:: :class:`~chainer.links.DeconvolutionND` to manage the model parameters ``W`` and ``b``. .. seealso:: :func:`deconvolution_2d` .. admonition:: Example **Example1**: the case when ``outsize`` is not given. >>> n = 10 >>> c_i, c_o = 3, 1 >>> d1, d2, d3 = 5, 10, 15 >>> k1, k2, k3 = 10, 10, 10 >>> p1, p2, p3 = 5, 5, 5 >>> x = np.random.uniform(0, 1, (n, c_i, d1, d2, d3)).\ astype(np.float32) >>> x.shape (10, 3, 5, 10, 15) >>> W = np.random.uniform(0, 1, (c_i, c_o, k1, k2, k3)).\ astype(np.float32) >>> W.shape (3, 1, 10, 10, 10) >>> b = np.random.uniform(0, 1, (c_o)).astype(np.float32) >>> b.shape (1,) >>> s1, s2, s3 = 2, 4, 6 >>> y = F.deconvolution_nd(x, W, b, stride=(s1, s2, s3), \ pad=(p1, p2, p3)) >>> y.shape (10, 1, 8, 36, 84) >>> l1 = s1 * (d1 - 1) + k1 - 2 * p1 >>> l2 = s2 * (d2 - 1) + k2 - 2 * p2 >>> l3 = s3 * (d3 - 1) + k3 - 2 * p3 >>> y.shape == (n, c_o, l1, l2, l3) True **Example2**: the case when ``outsize`` is given. >>> n = 10 >>> c_i, c_o = 3, 1 >>> d1, d2, d3 = 5, 10, 15 >>> k1, k2, k3 = 10, 10, 10 >>> p1, p2, p3 = 5, 5, 5 >>> x = np.random.uniform(0, 1, (n, c_i, d1, d2, d3)).\ astype(np.float32) >>> x.shape (10, 3, 5, 10, 15) >>> W = np.random.uniform(0, 1, (c_i, c_o, k1, k2, k3)).\ astype(np.float32) >>> W.shape (3, 1, 10, 10, 10) >>> b = np.random.uniform(0, 1, (c_o)).astype(np.float32) >>> b.shape (1,) >>> s1, s2, s3 = 2, 4, 6 >>> l1, l2, l3 = 9, 38, 87 >>> d1 == int((l1 + 2 * p1 - k1) / s1) + 1 True >>> d2 == int((l2 + 2 * p2 - k2) / s2) + 1 True >>> d3 == int((l3 + 2 * p3 - k3) / s3) + 1 True >>> y = F.deconvolution_nd(x, W, b, stride=(s1, s2, s3), \ pad=(p1, p2, p3), outsize=(l1, l2, l3)) >>> y.shape (10, 1, 9, 38, 87) >>> y.shape == (n, c_o, l1, l2, l3) True """ ndim = len(x.shape[2:]) func = DeconvolutionND( ndim, stride, pad, outsize, dilate=dilate, groups=groups) args = (x, W) if b is None else (x, W, b) y, = func.apply(args) return y def deconvolution_1d(x, W, b=None, stride=1, pad=0, outsize=None, dilate=1, groups=1): """1-dimensional deconvolution function. .. note:: This function calls :func:`~chainer.functions.deconvolution_nd` internally, so see the details of the behavior in the documentation of :func:`~chainer.functions.deconvolution_nd`. """ if len(x.shape[2:]) != 1: raise ValueError( 'The number of dimensions under channel dimension of the input ' '\'x\' should be 1. But the actual ndim was {}.'.format( len(x.shape[2:]))) return deconvolution_nd(x, W, b, stride, pad, outsize, dilate, groups) def deconvolution_3d(x, W, b=None, stride=1, pad=0, outsize=None, dilate=1, groups=1): """3-dimensional deconvolution function. .. note:: This function calls :func:`~chainer.functions.deconvolution_nd` internally, so see the details of the behavior in the documentation of :func:`~chainer.functions.deconvolution_nd`. """ if len(x.shape[2:]) != 3: raise ValueError( 'The number of dimensions under channel dimension of the input ' '\'x\' should be 3. But the actual ndim was {}.'.format( len(x.shape[2:]))) return deconvolution_nd(x, W, b, stride, pad, outsize, dilate, groups)
16,246
35.428251
79
py
chainer
chainer-master/chainer/functions/connection/depthwise_convolution_2d.py
import chainer def depthwise_convolution_2d(x, W, b=None, stride=1, pad=0): """Two-dimensional depthwise convolution function. This is an implementation of two-dimensional depthwise convolution. It takes two or three variables: the input image ``x``, the filter weight ``W``, and optionally, the bias vector ``b``. Notation: here is a notation for dimensionalities. - :math:`n` is the batch size. - :math:`c_I` is the number of the input. - :math:`c_M` is the channel multiplier. - :math:`h` and :math:`w` are the height and width of the input image, respectively. - :math:`h_O` and :math:`w_O` are the height and width of the output image, respectively. - :math:`k_H` and :math:`k_W` are the height and width of the filters, respectively. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable of shape :math:`(n, c_I, h, w)`. W (:class:`~chainer.Variable` or :ref:`ndarray`): Weight variable of shape :math:`(c_M, c_I, k_H, k_W)`. b (:class:`~chainer.Variable` or :ref:`ndarray`): Bias variable of length :math:`c_M * c_I` (optional). stride (int or pair of ints): Stride of filter applications. ``stride=s`` and ``stride=(s, s)`` are equivalent. pad (int or pair of ints): Spatial padding width for input arrays. ``pad=p`` and ``pad=(p, p)`` are equivalent. Returns: ~chainer.Variable: Output variable. Its shape is :math:`(n, c_I * c_M, h_O, w_O)`. Like ``Convolution2D``, ``DepthwiseConvolution2D`` function computes correlations between filters and patches of size :math:`(k_H, k_W)` in ``x``. But unlike ``Convolution2D``, ``DepthwiseConvolution2D`` does not add up input channels of filters but concatenates them. For that reason, the shape of outputs of depthwise convolution are :math:`(n, c_I * c_M, h_O, w_O)`, :math:`c_M` is called channel_multiplier. :math:`(h_O, w_O)` is determined by the equivalent equation of ``Convolution2D``. If the bias vector is given, then it is added to all spatial locations of the output of convolution. See: `L. Sifre. Rigid-motion scattering for image classification <https://www.di.ens.fr/data/publications/papers/phd_sifre.pdf>`_ .. seealso:: :class:`~chainer.links.DepthwiseConvolution2D` to manage the model parameters ``W`` and ``b``. .. admonition:: Example >>> x = np.random.uniform(0, 1, (2, 3, 4, 7)) >>> W = np.random.uniform(0, 1, (2, 3, 3, 3)) >>> b = np.random.uniform(0, 1, (6,)) >>> y = F.depthwise_convolution_2d(x, W, b) >>> y.shape (2, 6, 2, 5) """ multiplier, in_channels, kh, kw = W.shape F = chainer.functions W = F.transpose(W, (1, 0, 2, 3)) W = F.reshape(W, (multiplier * in_channels, 1, kh, kw)) return F.convolution_2d(x, W, b, stride, pad, groups=in_channels)
3,003
38.012987
79
py
chainer
chainer-master/chainer/functions/connection/local_convolution_2d.py
from six import moves import chainer from chainer import backend from chainer import function_node from chainer.utils import type_check from chainer import variable def _pair(x): if hasattr(x, '__getitem__'): return x return x, x class LocalConvolution2DFunction(function_node.FunctionNode): def __init__(self, stride=1): self.sy, self.sx = _pair(stride) def check_type_forward(self, in_types): n_in = in_types.size() type_check.expect(2 <= n_in, n_in <= 3) x_type, w_type = in_types[:2] type_check.expect( x_type.dtype.kind == 'f', w_type.dtype.kind == 'f', x_type.ndim == 4, w_type.ndim == 6, x_type.shape[1] == w_type.shape[3], ) if type_check.eval(n_in) == 3: b_type = in_types[2] type_check.expect( b_type.dtype == x_type.dtype, b_type.ndim == 3, b_type.shape == w_type.shape[:3] ) def forward(self, inputs): # Channels-first is Chainer's tensor format # W is 6-dimensional x, W = inputs[:2] b = inputs[2] if len(inputs) == 3 else None stride_row, stride_col = self.sy, self.sx output_row, output_col = W.shape[1], W.shape[2] feature_dim = W.shape[3] * W.shape[4] * W.shape[5] xp = backend.get_array_module(*inputs) output = xp.empty((x.shape[0], W.shape[0], output_row, output_col,), dtype=x.dtype) for i in moves.range(output_row): for j in moves.range(output_col): slice_row = slice(i * stride_row, i * stride_row + W.shape[4]) slice_col = slice(j * stride_col, j * stride_col + W.shape[5]) x_flatten = xp.reshape(x[..., slice_row, slice_col], (-1, feature_dim)) W_flatten = xp.reshape(W[:, i, j, ...], (-1, feature_dim)) output[..., i, j] = xp.dot(x_flatten, W_flatten.T) if b is not None: output += b[None, :, :, :] self.retain_inputs((0, 1)) # only retain x and W return output, def backward(self, indices, grad_outputs): xvar, Wvar = self.get_retained_inputs() x = xvar.data W = Wvar.data gyvar, = grad_outputs gy = gyvar.data xp = backend.get_array_module(x, W) stride_row, stride_col = self.sy, self.sx output_row, output_col = W.shape[1], W.shape[2] ret = [] if 0 in indices: gx = xp.zeros_like(x) for i in moves.range(output_row): for j in moves.range(output_col): slice_row = slice(i * stride_row, i * stride_row + W.shape[4]) slice_col = slice(j * stride_col, j * stride_col + W.shape[5]) # ochans * ichans * krows * kcols W_slice = W[:, i, j, ...] # nsamps * ochans gy_slice = gy[..., i, j] # -> nsamps * ichans * krows * kcols gx[:, :, slice_row, slice_col] += xp.tensordot( gy_slice, W_slice, axes=[(1,), (0,)] ) ret.append(chainer.functions.cast(variable.as_variable(gx), x.dtype)) if 1 in indices: gW = xp.empty_like(W) for i in moves.range(output_row): for j in moves.range(output_col): slice_row = slice(i * stride_row, i * stride_row + W.shape[4]) slice_col = slice(j * stride_col, j * stride_col + W.shape[5]) # nsamps * inchans * krows * kcols x_slice = x[:, :, slice_row, slice_col] # nsamps * outchans * 1 * 1 gy_slice = gy[:, :, i, j] gW[:, i, j, :, :, :] = xp.tensordot( gy_slice, x_slice, axes=[(0,), (0,)] ) ret.append(chainer.functions.cast(variable.as_variable(gW), W.dtype)) if 2 in indices: gb = chainer.functions.sum(gyvar, axis=0) ret.append(gb) return ret def local_convolution_2d(x, W, b=None, stride=1): """Two-dimensional local convolution function. Locally-connected function for 2D inputs. Works similarly to convolution_2d, except that weights are unshared, that is, a different set of filters is applied at each different patch of the input. It takes two or three variables: the input image ``x``, the filter weight ``W``, and optionally, the bias vector ``b``. Notation: here is a notation for dimensionalities. - :math:`n` is the batch size. - :math:`c_I` is the number of the input. - :math:`c_O` is the number of output channels. - :math:`h` and :math:`w` are the height and width of the input image, respectively. - :math:`h_O` and :math:`w_O` are the height and width of the output image, respectively. - :math:`k_H` and :math:`k_W` are the height and width of the filters, respectively. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable of shape :math:`(n, c_I, h, w)`. W (:class:`~chainer.Variable` or :ref:`ndarray`): Weight variable of shape :math:`(c_O, h_O, w_O, c_I, k_H, k_W)`. b (:class:`~chainer.Variable` or :ref:`ndarray`): Bias variable of shape :math:`(c_O,h_O,w_O)` (optional). stride (int or pair of ints): Stride of filter applications. ``stride=s`` and ``stride=(s, s)`` are equivalent. Returns: ~chainer.Variable: Output variable. Its shape is :math:`(n, c_O, h_O, w_O)`. Like ``Convolution2D``, ``LocalConvolution2D`` function computes correlations between filters and patches of size :math:`(k_H, k_W)` in ``x``. But unlike ``Convolution2D``, ``LocalConvolution2D`` has a separate filter for each patch of the input :math:`(h_O, w_O)` is determined by the equivalent equation of ``Convolution2D``, without any padding If the bias vector is given, then it is added to all spatial locations of the output of convolution. .. seealso:: :class:`~chainer.links.LocalConvolution2D` to manage the model parameters ``W`` and ``b``. .. admonition:: Example >>> x = np.random.uniform(0, 1, (2, 3, 7, 7)) >>> W = np.random.uniform(0, 1, (2, 5, 5, 3, 3, 3)) >>> b = np.random.uniform(0, 1, (2, 5, 5)) >>> y = F.local_convolution_2d(x, W, b) >>> y.shape (2, 2, 5, 5) """ fnode = LocalConvolution2DFunction(stride) if b is None: args = (x, W) else: args = (x, W, b) y, = fnode.apply(args) return y
7,212
36.373057
79
py
chainer
chainer-master/chainer/functions/connection/__init__.py
0
0
0
py
chainer
chainer-master/chainer/functions/connection/convolution_2d.py
import numpy import chainer from chainer import backend from chainer.backends import cuda from chainer.backends import intel64 from chainer import configuration from chainer import function_node import chainer.functions from chainer import memory_layouts from chainer.utils import argument from chainer.utils import conv from chainer.utils import type_check import chainerx if cuda.cudnn_enabled: _cudnn_version = cuda.cuda.cudnn.getVersion() # type: ignore def _pair(x): if hasattr(x, '__getitem__'): return x return x, x # Used by deconvolution_2d.py. # TODO(beam2d): Unify matmul implementations def _matmul(a, b): xp = backend.get_array_module(a) if not hasattr(xp, 'matmul'): # NumPy 1.9 does not support matmul. We use einsum instead. return xp.einsum('ijl,ilk->ijk', a, b) return xp.matmul(a, b) class Convolution2DFunction(function_node.FunctionNode): _use_ideep = False def __init__(self, stride=1, pad=0, cover_all=False, **kwargs): dilate, groups, cudnn_fast = argument.parse_kwargs( kwargs, ('dilate', 1), ('groups', 1), ('cudnn_fast', False), deterministic='deterministic argument is not supported anymore. ' 'Use chainer.using_config(\'cudnn_deterministic\', value) context ' 'where value is either `True` or `False`.', requires_x_grad='requires_x_grad argument is not supported ' 'anymore. Just remove the argument. Note that whether to compute ' 'the gradient w.r.t. x is automatically decided during ' 'backpropagation.') self.sy, self.sx = _pair(stride) self.ph, self.pw = _pair(pad) self.cover_all = cover_all self.dy, self.dx = _pair(dilate) self.groups = groups self.cudnn_fast = cudnn_fast if self.dx < 1 or self.dy < 1: raise ValueError('Dilate should be positive, but {} is ' 'supplied.'.format(dilate)) def check_type_forward(self, in_types): n_in = in_types.size() type_check.expect(2 <= n_in, n_in <= 3) x_type = in_types[0] w_type = in_types[1] type_check.expect( x_type.dtype.kind == 'f', w_type.dtype.kind == 'f', x_type.ndim == 4, w_type.ndim == 4, x_type.shape[1] == w_type.shape[1] * self.groups, ) if type_check.eval(n_in) == 3: b_type = in_types[2] type_check.expect( b_type.dtype == x_type.dtype, b_type.ndim == 1, b_type.shape[0] == w_type.shape[0], ) def check_layout_forward(self, inputs): input_layouts = self.input_layouts n = len(inputs) layouts = ( (memory_layouts.CUDNN_CHANNEL_FIRST_X, memory_layouts.CUDNN_CHANNEL_LAST_X), (memory_layouts.CUDNN_CHANNEL_FIRST_W, memory_layouts.CUDNN_CHANNEL_LAST_W), (None,)) for i, (input_layout, expected_layouts) in ( enumerate(zip(input_layouts, layouts[:n]))): if input_layout not in expected_layouts: raise RuntimeError( 'Invalid layout for input {}: {}'.format(i, input_layout)) def _get_out_size(self, x_shape, w_shape): _, _, kh, kw = w_shape _, _, h, w = x_shape out_h = conv.get_conv_outsize( h, kh, self.sy, self.ph, cover_all=self.cover_all, d=self.dy) if out_h <= 0: raise RuntimeError('Height in the output should be positive.') out_w = conv.get_conv_outsize( w, kw, self.sx, self.pw, cover_all=self.cover_all, d=self.dx) if out_w <= 0: raise RuntimeError('Width in the output should be positive.') return out_h, out_w def _check_input_layouts_all_standard(self): if not all([layout is None for layout in self.input_layouts]): raise RuntimeError( 'Non-standard memory layouts are only supported with cupy ' 'arrays in {}. Input layouts: {}'.format( self.label, self.input_layouts)) def forward_chainerx(self, inputs): # TODO(hvy): Support mixed precision. if any([arr.dtype != inputs[0].dtype for arr in inputs[1:]]): return chainer.Fallback # TODO(hvy): Support dilate > 1. if self.dy > 1 or self.dx > 1: return chainer.Fallback # TODO(hvy): Support groups > 1. if self.groups > 1: return chainer.Fallback if inputs[0].device.backend.name == 'cuda' and self.cover_all: return chainer.Fallback return chainerx.conv( *inputs, stride=(self.sy, self.sx), pad=(self.ph, self.pw), cover_all=self.cover_all), def forward_cpu(self, inputs): if self.cudnn_fast: raise RuntimeError( '\'cudnn_fast\' can\'t be used in the CPU backend') self._check_input_layouts_all_standard() self.retain_inputs((0, 1)) # retain only x and W if len(inputs) == 2: (x, W), b = inputs, None else: x, W, b = inputs if (intel64.should_use_ideep('>=auto') and intel64.inputs_all_ready(inputs)): self._use_ideep = True if self.groups > 1: return self._forward_grouped_convolution(x, W, b) else: return self._forward_cpu_core(x, W, b) def _forward_cpu_core(self, x, W, b): if self._use_ideep: return self._forward_ideep(x, W, b) kh, kw = W.shape[2:] col = conv.im2col_cpu( x, kh, kw, self.sy, self.sx, self.ph, self.pw, cover_all=self.cover_all, dy=self.dy, dx=self.dx) y = numpy.tensordot( col, W, ((1, 2, 3), (1, 2, 3))).astype(x.dtype, copy=False) if b is not None: y += b y = numpy.rollaxis(y, 3, 1) return y, def _forward_ideep(self, x, W, b): out_c, input_c, kh, kw = W.shape n, c, h, w = x.shape out_h, out_w = self._get_out_size(x.shape, W.shape) pd = (self.sy * (out_h - 1) + (kh + (kh - 1) * (self.dy - 1)) - h - self.ph) pr = (self.sx * (out_w - 1) + (kw + (kw - 1) * (self.dx - 1)) - w - self.pw) param = intel64.ideep.convolution2DParam( (n, out_c, out_h, out_w), self.dy, self.dx, self.sy, self.sx, self.ph, self.pw, pd, pr) y = intel64.ideep.convolution2D.Forward( intel64.ideep.array(x), intel64.ideep.array(W), intel64.ideep.array(b) if b is not None else None, param) return y, def forward_gpu(self, inputs): self.retain_inputs((0, 1)) # retain only x and W if len(inputs) == 2: (x, W), b = inputs, None x_layout, w_layout = self.input_layouts else: x, W, b = inputs x_layout, w_layout, _ = self.input_layouts x_shape = memory_layouts._transpose_shape(x.shape, x_layout, None) w_shape = memory_layouts._transpose_shape(W.shape, w_layout, None) n, _, h, w = x_shape out_c, _, kh, kw = w_shape out_h, out_w = self._get_out_size(x_shape, w_shape) y_raw_shape = memory_layouts._transpose_shape( (n, out_c, out_h, out_w), None, x_layout) y = cuda.cupy.empty(y_raw_shape, dtype=x.dtype) use_cudnn = ( chainer.should_use_cudnn('>=auto') and not self.cover_all and x.dtype == W.dtype and ((self.dy == 1 and self.dx == 1) or _cudnn_version >= 6000) and (self.groups <= 1 or _cudnn_version >= 7000) ) if self.cudnn_fast and not use_cudnn: raise RuntimeError('\'cudnn_fast\' requires cuDNN to work') if use_cudnn: # cuDNN implementation return self._forward_cudnn(x, W, b, y, (x_layout, w_layout)) elif self.groups > 1: return self._forward_grouped_convolution(x, W, b) else: return self._forward_gpu_core(x, W, b) def _forward_gpu_core(self, x, W, b): kh, kw = W.shape[2:] # Implementation using im2col col = conv.im2col_gpu( x, kh, kw, self.sy, self.sx, self.ph, self.pw, cover_all=self.cover_all, dy=self.dy, dx=self.dx) y = cuda.cupy.tensordot( col, W, ((1, 2, 3), (1, 2, 3))).astype(x.dtype, copy=False) # TODO(beam2d): Support unshared bias if b is not None: y += b y = cuda.cupy.rollaxis(y, 3, 1) return y, def _forward_grouped_convolution(self, x, W, b): # G: group count # N: batch size # kH, kW: kernel height, kernel width # iC, iH, iW: input channels, input height, input width # oC, oH, oW: output channels, output height, output width G = self.groups N, iC, iH, iW = x.shape oC, _, kH, kW = W.shape # _ == iCg iCg = iC // G oCg = oC // G # (N, iC, kW, kW, oH, oW) x = conv.im2col(x, kH, kW, self.sy, self.sx, self.ph, self.pw, cover_all=self.cover_all, dy=self.dy, dx=self.dx) oH, oW = x.shape[-2:] x = x.transpose(1, 2, 3, 0, 4, 5) # (iC, kH, kW, N, oH, oW) x = x.reshape(G, iCg * kH * kW, N * oH * oW) W = W.reshape(G, oCg, iCg * kH * kW) # (G, oCg, N*oH*oW) = (G, oCg, iCg*kH*kW) @ (G, iCg*kH*kW, N*oH*oW) y = _matmul(W, x).astype(x.dtype, copy=False) y = y.reshape(oC, N, oH, oW) y = y.transpose(1, 0, 2, 3) # (N, oC, oH, oW) if b is not None: y += b.reshape(1, b.size, 1, 1) return y, def _forward_cudnn(self, x, W, b, y, input_layouts): x_layout, w_layout = input_layouts self.output_layouts = (x_layout,) pad = (self.ph, self.pw) stride = (self.sy, self.sx) dilation = (self.dy, self.dx) auto_tune = configuration.config.autotune tensor_core = configuration.config.use_cudnn_tensor_core cudnn_x_layout = cuda._get_cudnn_tensor_layout_x(x_layout) cudnn_w_layout = cuda._get_cudnn_tensor_layout_w(w_layout) cuda.cudnn.convolution_forward( x, W, b, y, pad, stride, dilation, self.groups, auto_tune=auto_tune, tensor_core=tensor_core, d_layout=cudnn_x_layout, w_layout=cudnn_w_layout) return y, def backward(self, indexes, grad_outputs): x, W = self.get_retained_inputs() if len(self.input_layouts) == 2: x_layout, _ = self.input_layouts else: x_layout, _, _ = self.input_layouts gy, = grad_outputs ret = [] if 0 in indexes: _, _, xh, xw = x.shape gx = chainer.functions.deconvolution_2d( gy, W, stride=(self.sy, self.sx), pad=(self.ph, self.pw), outsize=(xh, xw), dilate=(self.dy, self.dx), groups=self.groups) assert gx.shape == x.shape ret.append(gx) if 1 in indexes: gW, = Convolution2DGradW( self, W.shape, W.dtype, W.layout).apply((x, gy)) ret.append(gW) if 2 in indexes: axis = (0, 2, 3) inv_trans = memory_layouts._get_layout_transpose_axes( gy.ndim, None, x_layout) if inv_trans is None: raw_axis = axis else: raw_axis = tuple([inv_trans[i] for i in axis]) gb = chainer.functions.sum(gy, axis=raw_axis) ret.append(gb) return ret class Convolution2DGradW(function_node.FunctionNode): def __init__(self, conv2d, w_shape, w_dtype, w_layout): self.kh, self.kw = w_shape[2::] self.sy = conv2d.sy self.sx = conv2d.sx self.ph = conv2d.ph self.pw = conv2d.pw self.dy = conv2d.dy self.dx = conv2d.dx self.cover_all = conv2d.cover_all self.W_shape = w_shape self.W_dtype = w_dtype self.w_layout = w_layout self.groups = conv2d.groups self._use_ideep = conv2d._use_ideep def check_layout_forward(self, inputs): pass def forward_cpu(self, inputs): self.retain_inputs((0, 1)) x, gy = inputs if self.groups > 1: return self._forward_grouped_convolution(x, gy) else: return self._forward_cpu_core(x, gy) def _forward_cpu_core(self, x, gy): if self._use_ideep: return self._forward_ideep(x, gy) # NumPy raises an error when the array is not contiguous. # See: https://github.com/chainer/chainer/issues/2744 # TODO(niboshi): Remove this code when NumPy is fixed. if (not (gy.flags.c_contiguous or gy.flags.f_contiguous) and 1 in gy.shape): gy = numpy.ascontiguousarray(gy) col = conv.im2col_cpu( x, self.kh, self.kw, self.sy, self.sx, self.ph, self.pw, cover_all=self.cover_all, dy=self.dy, dx=self.dx) gW = numpy.tensordot(gy, col, ((0, 2, 3), (0, 4, 5)) ).astype(self.W_dtype, copy=False) return gW, def _forward_ideep(self, x, gy): n, input_c, h, w = x.shape n, out_c, out_h, out_w = gy.shape pd = (self.sy * (out_h - 1) + (self.kh + (self.kh - 1) * (self.dy - 1)) - h - self.ph) pr = (self.sx * (out_w - 1) + (self.kw + (self.kw - 1) * (self.dx - 1)) - w - self.pw) param = intel64.ideep.convolution2DParam( (out_c, input_c, self.kh, self.kw), self.dy, self.dx, self.sy, self.sx, self.ph, self.pw, pd, pr) gW = intel64.ideep.convolution2D.BackwardWeights( intel64.ideep.array(x), intel64.ideep.array(gy), param) return gW, def forward_gpu(self, inputs): self.retain_inputs((0, 1)) x, gy = inputs use_cudnn = ( chainer.should_use_cudnn('>=auto') and not self.cover_all and x.dtype == self.W_dtype and ((self.dy == 1 and self.dx == 1) or (_cudnn_version >= 6000 and not configuration.config.cudnn_deterministic)) and (self.groups <= 1 or _cudnn_version >= 7000) ) if use_cudnn: # cuDNN implementation return self._forward_cudnn(x, gy) elif self.groups > 1: return self._forward_grouped_convolution(x, gy) else: return self._forward_gpu_core(x, gy) def _forward_gpu_core(self, x, gy): col = conv.im2col_gpu( x, self.kh, self.kw, self.sy, self.sx, self.ph, self.pw, cover_all=self.cover_all, dy=self.dy, dx=self.dx) gW = cuda.cupy.tensordot(gy, col, ((0, 2, 3), (0, 4, 5)) ).astype(self.W_dtype, copy=False) return gW, def _forward_grouped_convolution(self, x, gy): # G: group count # N: batch size # kH, kW: kernel height, kernel width # iC, iH, iW: input channels, input height, input width # oC, oH, oW: output channels, output height, output width G = self.groups N, iC, iH, iW = x.shape _, oC, oH, oW = gy.shape # _ == N kH = self.kh kW = self.kw iCg = iC // G oCg = oC // G # (N, iC, kH, kW, oH, oW) x = conv.im2col(x, kH, kW, self.sy, self.sx, self.ph, self.pw, cover_all=self.cover_all, dy=self.dy, dx=self.dx) x = x.transpose(1, 2, 3, 0, 4, 5) # (iC, kH, kW, N, oH, oW) x = x.reshape(G, iCg * kH * kW, N * oH * oW) x = x.transpose(0, 2, 1) # (G, N*oH*oW, iCg*kH*kW) gy = gy.transpose(1, 0, 2, 3) # (oC, N, oH, oW) gy = gy.reshape(G, oCg, N * oH * oW) # (G, oCg, iCg*kH*kW) = (G, oCg, N*oH*oW) @ (G, N*oH*oW, iCg*kH*kW) gW = _matmul(gy, x).astype(self.W_dtype, copy=False) gW = gW.reshape(oC, iCg, kH, kW) return gW, def _forward_cudnn(self, x, gy): x_layout, gy_layout = self.input_layouts w_layout = self.w_layout w_raw_shape = memory_layouts._transpose_shape( self.W_shape, None, w_layout) gW = cuda.cupy.empty(w_raw_shape, dtype=self.W_dtype) pad = (self.ph, self.pw) stride = (self.sy, self.sx) dilation = (self.dy, self.dx) deterministic = configuration.config.cudnn_deterministic auto_tune = configuration.config.autotune tensor_core = configuration.config.use_cudnn_tensor_core cudnn_x_layout = cuda._get_cudnn_tensor_layout_x(x_layout) cudnn_w_layout = cuda._get_cudnn_tensor_layout_w(w_layout) cuda.cudnn.convolution_backward_filter( x, gy, gW, pad, stride, dilation, self.groups, deterministic=deterministic, auto_tune=auto_tune, tensor_core=tensor_core, d_layout=cudnn_x_layout, w_layout=cudnn_w_layout) return gW, def backward(self, indexes, grad_outputs): x, gy = self.get_retained_inputs() ggW, = grad_outputs ret = [] if 0 in indexes: xh, xw = x.shape[2:] gx = chainer.functions.deconvolution_2d( gy, ggW, stride=(self.sy, self.sx), pad=(self.ph, self.pw), outsize=(xh, xw), dilate=(self.dy, self.dx), groups=self.groups) ret.append(gx) if 1 in indexes: ggy = convolution_2d( x, ggW, stride=(self.sy, self.sx), pad=(self.ph, self.pw), cover_all=self.cover_all, dilate=(self.dy, self.dx), groups=self.groups) ret.append(ggy) return ret def convolution_2d(x, W, b=None, stride=1, pad=0, cover_all=False, **kwargs): """convolution_2d(x, W, b=None, stride=1, pad=0, cover_all=False, *, \ dilate=1, groups=1) Two-dimensional convolution function. This is an implementation of two-dimensional convolution in ConvNets. It takes three variables: the input image ``x``, the filter weight ``W``, and the bias vector ``b``. Notation: here is a notation for dimensionalities. - :math:`n` is the batch size. - :math:`c_I` and :math:`c_O` are the number of the input and output channels, respectively. - :math:`h_I` and :math:`w_I` are the height and width of the input image, respectively. - :math:`h_K` and :math:`w_K` are the height and width of the filters, respectively. - :math:`h_P` and :math:`w_P` are the height and width of the spatial padding size, respectively. Then the ``Convolution2D`` function computes correlations between filters and patches of size :math:`(h_K, w_K)` in ``x``. Note that correlation here is equivalent to the inner product between expanded vectors. Patches are extracted at positions shifted by multiples of ``stride`` from the first position ``(-h_P, -w_P)`` for each spatial axis. The right-most (or bottom-most) patches do not run over the padded spatial size. Let :math:`(s_Y, s_X)` be the stride of filter application. Then, the output size :math:`(h_O, w_O)` is determined by the following equations: .. math:: h_O &= (h_I + 2h_P - h_K) / s_Y + 1,\\\\ w_O &= (w_I + 2w_P - w_K) / s_X + 1. If ``cover_all`` option is ``True``, the filter will cover the all spatial locations. So, if the last stride of filter does not cover the end of spatial locations, an additional stride will be applied to the end part of spatial locations. In this case, the output size :math:`(h_O, w_O)` is determined by the following equations: .. math:: h_O &= (h_I + 2h_P - h_K + s_Y - 1) / s_Y + 1,\\\\ w_O &= (w_I + 2w_P - w_K + s_X - 1) / s_X + 1. If the bias vector is given, then it is added to all spatial locations of the output of convolution. The output of this function can be non-deterministic when it uses cuDNN. If ``chainer.configuration.config.cudnn_deterministic`` is ``True`` and cuDNN version is >= v3, it forces cuDNN to use a deterministic algorithm. Convolution links can use a feature of cuDNN called autotuning, which selects the most efficient CNN algorithm for images of fixed-size, can provide a significant performance boost for fixed neural nets. To enable, set `chainer.using_config('autotune', True)` When the dilation factor is greater than one, cuDNN is not used unless the version is 6.0 or higher. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable of shape :math:`(n, c_I, h_I, w_I)`. W (:class:`~chainer.Variable` or :ref:`ndarray`): Weight variable of shape :math:`(c_O, c_I, h_K, w_K)`. b (None or :class:`~chainer.Variable` or :ref:`ndarray`): Bias variable of length :math:`c_O` (optional). stride (:class:`int` or pair of :class:`int` s): Stride of filter applications. ``stride=s`` and ``stride=(s, s)`` are equivalent. pad (:class:`int` or pair of :class:`int` s): Spatial padding width for input arrays. ``pad=p`` and ``pad=(p, p)`` are equivalent. cover_all (:class:`bool`): If ``True``, all spatial locations are convoluted into some output pixels. dilate (:class:`int` or pair of :class:`int` s): Dilation factor of filter applications. ``dilate=d`` and ``dilate=(d, d)`` are equivalent. groups (:class:`int`): Number of groups of channels. If the number is greater than 1, input tensor :math:`W` is divided into some blocks by this value. For each tensor blocks, convolution operation will be executed independently. Input channel size :math:`c_I` and output channel size :math:`c_O` must be exactly divisible by this value. Returns: ~chainer.Variable: Output variable of shape :math:`(n, c_O, h_O, w_O)`. .. seealso:: :class:`~chainer.links.Convolution2D` to manage the model parameters ``W`` and ``b``. .. admonition:: Example >>> n = 10 >>> c_i, c_o = 3, 1 >>> h_i, w_i = 30, 40 >>> h_k, w_k = 10, 10 >>> h_p, w_p = 5, 5 >>> x = np.random.uniform(0, 1, (n, c_i, h_i, w_i)).astype(np.float32) >>> x.shape (10, 3, 30, 40) >>> W = np.random.uniform(0, 1, (c_o, c_i, h_k, w_k)).\ astype(np.float32) >>> W.shape (1, 3, 10, 10) >>> b = np.random.uniform(0, 1, (c_o,)).astype(np.float32) >>> b.shape (1,) >>> s_y, s_x = 5, 7 >>> y = F.convolution_2d(x, W, b, stride=(s_y, s_x), pad=(h_p, w_p)) >>> y.shape (10, 1, 7, 6) >>> h_o = int((h_i + 2 * h_p - h_k) / s_y + 1) >>> w_o = int((w_i + 2 * w_p - w_k) / s_x + 1) >>> y.shape == (n, c_o, h_o, w_o) True >>> y = F.convolution_2d(x, W, b, stride=(s_y, s_x), pad=(h_p, w_p), \ cover_all=True) >>> y.shape == (n, c_o, h_o, w_o + 1) True """ dilate, groups, cudnn_fast = argument.parse_kwargs( kwargs, ('dilate', 1), ('groups', 1), ('cudnn_fast', False), deterministic='deterministic argument is not supported anymore. ' 'Use chainer.using_config(\'cudnn_deterministic\', value) ' 'context where value is either `True` or `False`.') fnode = Convolution2DFunction(stride, pad, cover_all, dilate=dilate, groups=groups, cudnn_fast=cudnn_fast) if b is None: args = x, W else: args = x, W, b y, = fnode.apply(args) return y
24,207
35.678788
79
py
chainer
chainer-master/chainer/functions/connection/embed_id.py
import numpy import six import chainer from chainer import backend from chainer.backends import cuda from chainer import function_node from chainer import utils from chainer.utils import type_check class EmbedIDFunction(function_node.FunctionNode): def __init__(self, ignore_label=None): self.ignore_label = ignore_label def check_type_forward(self, in_types): type_check.expect(in_types.size() == 2) x_type, w_type = in_types type_check.expect( x_type.dtype.kind == 'i', x_type.ndim >= 1, ) type_check.expect( w_type.dtype.kind == 'f', w_type.ndim == 2 ) def forward(self, inputs): self.retain_inputs((0,)) x, W = inputs self._w_shape = W.shape xp = backend.get_array_module(*inputs) if chainer.is_debug(): valid_x = xp.logical_and(0 <= x, x < len(W)) if self.ignore_label is not None: valid_x = xp.logical_or(valid_x, x == self.ignore_label) if not valid_x.all(): raise ValueError('Each not ignored `x` value need to satisfy ' '`0 <= x < len(W)`') if self.ignore_label is not None: mask = (x == self.ignore_label) return xp.where(mask[..., None], 0, W[xp.where(mask, 0, x)]), return W[x], def backward(self, indexes, grad_outputs): inputs = self.get_retained_inputs() gW = EmbedIDGrad( self._w_shape, self.ignore_label).apply(inputs + grad_outputs)[0] return None, gW class EmbedIDGrad(function_node.FunctionNode): def __init__(self, w_shape, ignore_label=None): self.w_shape = w_shape self.ignore_label = ignore_label def forward(self, inputs): self.retain_inputs((0,)) xp = backend.get_array_module(*inputs) x, gy = inputs self._gy_shape = gy.shape gW = xp.zeros(self.w_shape, dtype=gy.dtype) if xp is numpy: # It is equivalent to `numpy.add.at(gW, x, gy)` but ufunc.at is # too slow. for ix, igy in six.moves.zip(x.ravel(), gy.reshape(x.size, -1)): if ix == self.ignore_label: continue gW[ix] += igy else: utils.nondeterministic('atomicAdd') if self.ignore_label is None: cuda.elementwise( 'T gy, S x, S n_out', 'raw T gW', 'ptrdiff_t w_ind[] = {x, i % n_out};' 'atomicAdd(&gW[w_ind], gy)', 'embed_id_bwd')( gy, xp.expand_dims(x, -1), gW.shape[1], gW) else: cuda.elementwise( 'T gy, S x, S n_out, S ignore', 'raw T gW', ''' if (x != ignore) { ptrdiff_t w_ind[] = {x, i % n_out}; atomicAdd(&gW[w_ind], gy); } ''', 'embed_id_bwd_ignore_label')( gy, xp.expand_dims(x, -1), gW.shape[1], self.ignore_label, gW) return gW, def backward(self, indexes, grads): xp = backend.get_array_module(*grads) x = self.get_retained_inputs()[0].data ggW = grads[0] if self.ignore_label is not None: mask = x == self.ignore_label # To prevent index out of bounds, we need to check if ignore_label # is inside of W. if not (0 <= self.ignore_label < self.w_shape[1]): x = xp.where(mask, 0, x) ggy = ggW[x] if self.ignore_label is not None: mask, zero, _ = xp.broadcast_arrays( mask[..., None], xp.zeros((), ggy.dtype), ggy.data) ggy = chainer.functions.where(mask, zero, ggy) return None, ggy def embed_id(x, W, ignore_label=None): """Efficient linear function for one-hot input. This function implements so called *word embeddings*. It takes two arguments: a set of IDs (words) ``x`` in :math:`B` dimensional integer vector, and a set of all ID (word) embeddings ``W`` in :math:`V \\times d` float matrix. It outputs :math:`B \\times d` matrix whose ``i``-th row is the ``x[i]``-th row of ``W``. This function is only differentiable on the input ``W``. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Batch vectors of IDs. Each element must be signed integer. W (:class:`~chainer.Variable` or :ref:`ndarray`): Distributed representation of each ID (a.k.a. word embeddings). ignore_label (:class:`int` or :class:`None`): If ``ignore_label`` is an int value, ``i``-th row of return value is filled with ``0``. Returns: ~chainer.Variable: Output variable. .. seealso:: :class:`~chainer.links.EmbedID` to manage the model parameter ``W``. .. admonition:: Example >>> x = np.array([2, 1]).astype(np.int32) >>> x array([2, 1], dtype=int32) >>> W = np.array([[0, 0, 0], ... [1, 1, 1], ... [2, 2, 2]]).astype(np.float32) >>> W array([[0., 0., 0.], [1., 1., 1.], [2., 2., 2.]], dtype=float32) >>> F.embed_id(x, W).array array([[2., 2., 2.], [1., 1., 1.]], dtype=float32) >>> F.embed_id(x, W, ignore_label=1).array array([[2., 2., 2.], [0., 0., 0.]], dtype=float32) """ return EmbedIDFunction(ignore_label=ignore_label).apply((x, W))[0]
5,774
33.171598
78
py
chainer
chainer-master/chainer/functions/connection/dilated_convolution_2d.py
from chainer.functions.connection import convolution_2d def dilated_convolution_2d(x, W, b=None, stride=1, pad=0, dilate=1, cover_all=False): """Two-dimensional dilated convolution function. This is an implementation of two-dimensional dilated convolution in ConvNets. It takes three variables: the input image ``x``, the filter weight ``W``, and the bias vector ``b``. .. note:: You can also perform dilated convolution by passing ``dilate`` argument to :class:`chainer.functions.convolution_2d`. The functionality is the same. Notation: here is a notation for dimensionalities. - :math:`n` is the batch size. - :math:`c_I` and :math:`c_O` are the number of the input and output, respectively. - :math:`h` and :math:`w` are the height and width of the input image, respectively. - :math:`k_H` and :math:`k_W` are the height and width of the filters, respectively. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable of shape :math:`(n, c_I, h, w)`. W (:class:`~chainer.Variable` or :ref:`ndarray`): Weight variable of shape :math:`(c_O, c_I, k_H, k_W)`. b (:class:`~chainer.Variable` or :ref:`ndarray`): Bias variable of length :math:`c_O` (optional). stride (int or pair of ints): Stride of filter applications. ``stride=s`` and ``stride=(s, s)`` are equivalent. pad (int or pair of ints): Spatial padding width for input arrays. ``pad=p`` and ``pad=(p, p)`` are equivalent. dilate (int or pair of ints): Dilation factor of filter applications. ``dilate=d`` and ``dilate=(d, d)`` are equivalent. cover_all (bool): If ``True``, all spatial locations are convoluted into some output pixels. It may make the output size larger. Returns: ~chainer.Variable: Output variable. The two-dimensional dilated convolution function is defined as follows. Then the ``DilatedConvolution2D`` function computes correlations between filters and patches of size :math:`(k_H, k_W)` in ``x``. Patches here are extracted at intervals of the dilation factor. Note that correlation here is equivalent to the inner product between expanded vectors. Patches are extracted at intervals of the dilation factor and at positions shifted by multiples of ``stride`` from the first position ``-pad`` for each spatial axis. The right-most (or bottom-most) patches do not run over the padded spatial size. Let :math:`(s_Y, s_X)` be the stride of filter application, :math:`(p_H, p_W)` the spatial padding size, and :math:`(d_Y, d_X)` the dilation factor of filter application. Then, the output size :math:`(h_O, w_O)` is determined by the following equations: .. math:: h_O &= (h + 2p_H - k_H - (k_H - 1) * (d_Y - 1)) / s_Y + 1,\\\\ w_O &= (w + 2p_W - k_W - (k_W - 1) * (d_X - 1)) / s_X + 1. If the bias vector is given, then it is added to all spatial locations of the output of convolution. """ return convolution_2d.convolution_2d(x, W, b, stride, pad, cover_all, dilate=dilate)
3,279
43.324324
79
py
chainer
chainer-master/chainer/functions/loss/mean_absolute_error.py
import numpy import chainer from chainer import backend from chainer import function_node from chainer.utils import type_check def _get_intermediate_dtype(dtype): # Returns the dtype for intermediate calculation. # For float16 input, float32 is used. # Otherwise the same dtype as the parameter is used. if dtype == numpy.float16: return numpy.float32 return dtype class MeanAbsoluteError(function_node.FunctionNode): """Mean absolute error function.""" def check_type_forward(self, in_types): type_check._argname(in_types, ('x0', 'x1')) type_check.expect( in_types[0].dtype.kind == 'f', in_types[0].dtype == in_types[1].dtype, in_types[0].shape == in_types[1].shape ) def forward_cpu(self, inputs): x0, x1 = inputs self.diff = x0 - x1 orig_dtype = self.diff.dtype dtype = _get_intermediate_dtype(orig_dtype) diff = self.diff.ravel().astype(dtype, copy=False) return numpy.array(abs(diff).sum() / diff.size, dtype=orig_dtype), def forward_gpu(self, inputs): x0, x1 = inputs self.diff = x0 - x1 orig_dtype = self.diff.dtype dtype = _get_intermediate_dtype(orig_dtype) diff = self.diff.ravel().astype(dtype, copy=False) return (abs(diff).sum() / diff.dtype.type(diff.size)).astype( orig_dtype, copy=False), def backward(self, indexes, grad_outputs): gy, = grad_outputs coeff = gy * gy.data.dtype.type(1. / self.diff.size) coeff = chainer.functions.broadcast_to(coeff, self.diff.shape) gx0 = coeff * backend.get_array_module(gy.data).sign(self.diff) return gx0, -gx0 def mean_absolute_error(x0, x1): """Mean absolute error function. The function computes the mean absolute error between two variables. The mean is taken over the minibatch. Args ``x0`` and ``x1`` must have the same dimensions. This function first calculates the absolute value differences between the corresponding elements in x0 and x1, and then returns the mean of those differences. Args: x0 (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. x1 (:class:`~chainer.Variable` or :ref:`ndarray`): Input variable. Returns: ~chainer.Variable: A variable holding an array representing the mean absolute error of two inputs. .. admonition:: Example 1D array examples: >>> x = np.array([1, 2, 3]).astype(np.float32) >>> y = np.array([0, 0, 0]).astype(np.float32) >>> F.mean_absolute_error(x, y) variable(2.) >>> x = np.array([1, 2, 3, 4, 5, 6]).astype(np.float32) >>> y = np.array([7, 8, 9, 10, 11, 12]).astype(np.float32) >>> F.mean_absolute_error(x, y) variable(6.) 2D array example: In this example, there are 4 elements, and thus 4 errors >>> x = np.array([[1, 2], [3, 4]]).astype(np.float32) >>> y = np.array([[8, 8], [8, 8]]).astype(np.float32) >>> F.mean_absolute_error(x, y) variable(5.5) 3D array example: In this example, there are 8 elements, and thus 8 errors >>> x = np.reshape(np.array([1, 2, 3, 4, 5, 6, 7, 8]), (2, 2, 2)) >>> y = np.reshape(np.array([8, 8, 8, 8, 8, 8, 8, 8]), (2, 2, 2)) >>> x = x.astype(np.float32) >>> y = y.astype(np.float32) >>> F.mean_absolute_error(x, y) variable(3.5) """ return MeanAbsoluteError().apply((x0, x1))[0]
3,585
32.830189
76
py
chainer
chainer-master/chainer/functions/loss/ctc.py
import numpy import six import chainer from chainer import backend from chainer.backends import cuda from chainer import function from chainer import utils from chainer.utils import collections_abc from chainer.utils import type_check def _logsumexp(a, xp, axis=None): vmax = xp.amax(a, axis=axis, keepdims=True) if xp is numpy: vmax += xp.log(xp.sum(xp.exp(a - vmax), axis=axis, keepdims=True, dtype=a.dtype)) else: _logsumexp_impl = cuda.reduce( 'T x, T vmax', 'T y', 'exp(x - vmax)', 'a + b', 'y += log(a)', '0', 'logsumexp_impl') _logsumexp_impl(a, vmax, vmax, axis=axis, keepdims=True) return xp.squeeze(vmax, axis=axis) def _softmax(x, xp): val = xp.exp(x - xp.amax(x, axis=2, keepdims=True)) val /= xp.sum(val, axis=2, keepdims=True) return val def _label_to_path(labels, blank_symbol, xp): path = xp.full((len(labels), labels.shape[1] * 2 + 1), blank_symbol, dtype=numpy.int32) path[:, 1::2] = labels return path def _flip_path(path, path_length, xp): """Flips label sequence. This function rotates a label sequence and flips it. ``path[b, t]`` stores a label at time ``t`` in ``b``-th batch. The rotated matrix ``r`` is defined as ``r[b, t] = path[b, t + path_length[b]]`` .. :: a b c d . . a b c d d c b a . e f . . . -> . . . e f -> f e . . . g h i j k g h i j k k j i h g """ n_batch, n_label = path.shape rotate = (xp.arange(n_label) + path_length[:, None]) % n_label return path[xp.arange(n_batch, dtype=xp.int32)[:, None], rotate][:, ::-1] def _flip_label_probability(y, input_length, xp): """Flips a label probability matrix. This function rotates a label probability matrix and flips it. ``y[i, b, l]`` stores log probability of label ``l`` at ``i``-th input in ``b``-th batch. The rotated matrix ``r`` is defined as ``r[i, b, l] = y[i + input_length[b], b, l]`` """ seq, n_batch, n_vocab = y.shape rotate = (xp.arange(seq, dtype=xp.int32)[:, None] + input_length) % seq return y[ rotate[:, :, None], xp.arange(n_batch, dtype=xp.int32)[None, :, None], xp.arange(n_vocab, dtype=xp.int32)[None, None, :]][::-1] def _flip_path_probability(prob, input_length, path_length, xp): """Flips a path probability matrix. This function returns a path probability matrix and flips it. ``prob[i, b, t]`` stores log probability at ``i``-th input and at time ``t`` in a output sequence in ``b``-th batch. The rotated matrix ``r`` is defined as ``r[i, j, k] = prob[i + input_length[j], j, k + path_length[j]]`` """ seq, n_batch, n_label = prob.shape rotate_input = ((xp.arange(seq, dtype=xp.int32)[:, None] + input_length) % seq) rotate_label = ((xp.arange(n_label, dtype=xp.int32) + path_length[:, None]) % n_label) return prob[ rotate_input[:, :, None], xp.arange(n_batch, dtype=xp.int32)[None, :, None], rotate_label][::-1, :, ::-1] class ConnectionistTemporalClassification(function.Function): """The implementation of Connectionist Temporal Classfication loss functions. To make it usable for real-world cases, this class has two policies below. 1. This class computes forward and backward variables in the log domain. 2. This class applies the softmax function to inputs. The Backward values of CTC loss is often overflows. This is avoided by computing backward values before the activation function is applied. """ def __init__(self, blank_symbol, reduce='mean'): self.blank_symbol = blank_symbol # Lazily initialized in the first forward computation for dtype self.zero_padding = None if reduce not in ('mean', 'no'): raise ValueError( 'only \'mean\' and \'no\' are valid ' 'for \'reduce\', but \'%s\' is given' % reduce) self.reduce = reduce def check_type_forward(self, in_types): type_check._argname( in_types, ('input_length', 'label_length', 't', 'x')) input_length_type, label_length_type, t_type, x_type = in_types type_check.expect( input_length_type.dtype == numpy.int32, input_length_type.ndim == 1, label_length_type.dtype == numpy.int32, label_length_type.ndim == 1, t_type.ndim == 2, t_type.dtype == numpy.int32, x_type.ndim == 3, x_type.dtype.kind == 'f', ) n_batch = x_type.shape[1] type_check.expect( t_type.shape[0] == n_batch, input_length_type.shape[0] == n_batch, label_length_type.shape[0] == n_batch, ) def log_matrix(self, x, xp): if xp == numpy: res = numpy.ma.log(x).filled(fill_value=self.zero_padding) else: create_recurrence_relation = cuda.elementwise( 'T x, T e', 'T y', 'y = x == 0 ? e : (T)log(x)', 'create_recurrence_relation') res = create_recurrence_relation(x, self.zero_padding) return res.astype(x.dtype, copy=False) # path probability to label probability def label_probability(self, label_size, path, path_length, multiply_seq, xp): seq_length = len(multiply_seq) n_batch = len(path) dtype = multiply_seq.dtype ret = xp.zeros((seq_length, n_batch, label_size), dtype) if xp == numpy: for b in six.moves.range(len(path)): target_path = path[b, :path_length[b]] chars = {c for c in target_path} for c in chars: ret[:, b, c] = xp.sum( multiply_seq[:, b, 0:path_length[b]] [:, target_path == c], axis=1) else: utils.nondeterministic('atomicAdd') cuda.elementwise( 'T prob, I path, I path_length, I max_path_length', 'raw T cum_prob', ''' I t = i % max_path_length; if (t < path_length) { int n_batch = cum_prob.shape()[1]; I s = i / (max_path_length * n_batch); I b = (i - s * (max_path_length * n_batch)) / max_path_length; int ind[] = {s, b, path}; atomicAdd(&cum_prob[ind], prob); } ''', 'ctc_label_prob_sum' )(multiply_seq, path, path_length[:, None], path.shape[1], ret) return ret def _computes_transition( self, prev_prob, path, path_length, cum_prob, y): xp = backend.get_array_module(prev_prob) if xp == numpy: n_batch, max_path_length = path.shape mat = xp.full( (3, n_batch, max_path_length), self.zero_padding, y.dtype) mat[0, :, :] = prev_prob mat[1, :, 1:] = prev_prob[:, :-1] mat[2, :, 2:] = prev_prob[:, :-2] # disable transition between the same symbols # (including blank-to-blank) same_transition = (path[:, :-2] == path[:, 2:]) mat[2, :, 2:][same_transition] = self.zero_padding prob = _logsumexp(mat, xp, axis=0) outside = xp.arange(max_path_length) >= path_length[:, None] prob[outside] = self.zero_padding cum_prob += prob batch_index = xp.arange(n_batch, dtype=xp.int32) prob += y[batch_index[:, None], path] else: prob = xp.empty_like(prev_prob) cuda.elementwise( 'raw T prob, raw I path, I path_length, T zero, raw T y', 'T z, T cum_prob', ''' int length = prob.shape()[1]; int b = i / length; int t = i - b * length; if (t >= path_length) { z = zero; cum_prob += zero; return; } int ind1[] = {b, t}; int ind2[] = {b, t - 1}; int ind3[] = {b, t - 2}; T f1 = prob[ind1]; T f2 = (0 <= t - 1) ? prob[ind2] : zero; T f3 = (0 <= t - 2 && path[ind3] != path[ind1]) ? prob[ind3] : zero; // calculates log-sum-exp T m = max(f1, max(f2, f3)); z = m + log(exp(f1 - m) + exp(f2 - m) + exp(f3 - m)); cum_prob += z; int y_ind[] = {b, path[ind1]}; z += y[y_ind]; ''', 'ctc_transition' )(prev_prob, path, path_length[:, None], self.zero_padding, y, prob, cum_prob) return prob def calc_trans(self, yseq, input_length, label, label_length, path, path_length, xp): max_input_length, n_batch, n_unit = yseq.shape max_label_length = label.shape[1] max_path_length = path.shape[1] assert label.shape == (n_batch, max_label_length), label.shape assert path.shape == (n_batch, max_label_length * 2 + 1) forward_prob = xp.full( (n_batch, max_path_length), self.zero_padding, dtype=yseq.dtype) forward_prob[:, 0] = 0 backward_prob = forward_prob batch_index = xp.arange(n_batch, dtype=xp.int32) seq_index = xp.arange(len(yseq), dtype=xp.int32) prob = yseq[seq_index[:, None, None], batch_index[:, None], path] # forward computation. for i, y in enumerate(yseq): forward_prob = self._computes_transition( forward_prob, path, path_length, prob[i], y) r_path = _flip_path(path, path_length, xp) yseq_inv = _flip_label_probability(yseq, input_length, xp) prob = _flip_path_probability(prob, input_length, path_length, xp) for i, y_inv in enumerate(yseq_inv): backward_prob = self._computes_transition( backward_prob, r_path, path_length, prob[i], y_inv) return _flip_path_probability(prob, input_length, path_length, xp) def forward(self, inputs): xp = backend.get_array_module(inputs[0]) self.input_length, label_length, t, xs = inputs if self.zero_padding is None: if xs.dtype == numpy.float16: self.zero_padding = -10000.0 else: self.zero_padding = -10000000000.0 if chainer.is_debug(): assert len(xs) >= xp.max(self.input_length) assert t.shape[1] >= xp.max(label_length) self.path_length = 2 * label_length + 1 self.yseq = _softmax(xs, xp) log_yseq = self.log_matrix(self.yseq, xp) self.path = _label_to_path(t, self.blank_symbol, xp) self.prob_trans = self.calc_trans( log_yseq, self.input_length, t, label_length, self.path, self.path_length, xp) loss = -_logsumexp(self.prob_trans[0], xp, axis=1) if self.reduce == 'mean': loss = utils.force_array(xp.mean(loss)) return loss, def backward(self, inputs, grad_output): xp = backend.get_array_module(inputs[0]) batch_size = len(inputs[2]) total_probability = _logsumexp(self.prob_trans[0], xp, axis=1) label_prob = self.label_probability( self.yseq.shape[2], self.path, self.path_length, xp.exp(self.prob_trans - total_probability[:, None]), xp) self.yseq -= label_prob if self.reduce == 'mean': self.yseq *= grad_output[0] / batch_size else: self.yseq *= grad_output[0][..., None] # mask self.yseq *= ( xp.arange(len(self.yseq))[:, None] < self.input_length)[..., None] return None, None, None, self.yseq def connectionist_temporal_classification( x, t, blank_symbol, input_length=None, label_length=None, reduce='mean'): """Connectionist Temporal Classification loss function. Connectionist Temporal Classification(CTC) [Graves2006]_ is a loss function of sequence labeling where the alignment between the inputs and target is unknown. See also [Graves2012]_ The output is a variable whose value depends on the value of the option ``reduce``. If it is ``'no'``, it holds the samplewise loss values. If it is ``'mean'``, it takes the mean of loss values. Args: x (list or tuple of :class:`~chainer.Variable`): A list of unnormalized probabilities for labels. Each element of ``x``, ``x[i]`` is a :class:`~chainer.Variable` object, which has shape ``(B, V)``, where ``B`` is the batch size and ``V`` is the number of labels. The softmax of ``x[i]`` represents the probabilities of the labels at time ``i``. t (:class:`~chainer.Variable` or :ref:`ndarray`): A matrix including expected label sequences. Its shape is ``(B, M)``, where ``B`` is the batch size and ``M`` is the maximum length of the label sequences. All elements in ``t`` must be less than ``V``, the number of labels. blank_symbol (int): Index of blank_symbol. This value must be non-negative. input_length (:class:`~chainer.Variable` or :ref:`ndarray`): Length of sequence for each of mini batch ``x`` (optional). Its shape must be ``(B,)``. If the ``input_length`` is omitted or ``None``, it assumes that all of ``x`` is valid input. label_length (:class:`~chainer.Variable` or :ref:`ndarray`): Length of sequence for each of mini batch ``t`` (optional). Its shape must be ``(B,)``. If the ``label_length`` is omitted or ``None``, it assumes that all of ``t`` is valid input. reduce (str): Reduction option. Its value must be either ``'mean'`` or ``'no'``. Otherwise, :class:`ValueError` is raised. Returns: ~chainer.Variable: A variable holding a scalar value of the CTC loss. If ``reduce`` is ``'no'``, the output variable holds array whose shape is `(B,)` where `B` is the number of samples. If it is ``'mean'``, it holds a scalar. .. note:: You need to input ``x`` without applying to activation functions(e.g. softmax function), because this function applies softmax functions to ``x`` before calculating CTC loss to avoid numerical limitations. You also need to apply softmax function to forwarded values before you decode it. .. note:: This function is differentiable only by ``x``. .. note:: This function supports (batch, sequence, 1-dimensional input)-data. .. [Graves2006] Alex Graves, Santiago Fernandez, Faustino Gomez, Jurgen Schmidhuber, `Connectionist Temporal Classification: Labelling Unsegmented Sequence Data with Recurrent Neural Networks <ftp://ftp.idsia.ch/pub/juergen/icml2006.pdf>`_ .. [Graves2012] Alex Graves, `Supervised Sequence Labelling with Recurrent Neural Networks <https://www.cs.toronto.edu/~graves/preprint.pdf>`_ """ if not isinstance(x, collections_abc.Sequence): raise TypeError('x must be a list of Variables') if not isinstance(blank_symbol, int): raise TypeError('blank_symbol must be non-negative integer.') assert 0 <= blank_symbol < x[0].shape[1] # This implementation only supports 1-dimensional data. # TODO(jnishi): Support d(>1)-dimensional inputs. assert x[0].ndim == 2 xp = backend.get_array_module(x[0]) if input_length is None: input_length = xp.full(len(x[0]), len(x), dtype=numpy.int32) if label_length is None: label_length = xp.full(len(t), t.shape[1], dtype=numpy.int32) return ConnectionistTemporalClassification(blank_symbol, reduce)( input_length, label_length, t, chainer.functions.stack(x))
16,314
38.21875
81
py
chainer
chainer-master/chainer/functions/loss/absolute_error.py
from chainer import backend from chainer import function_node from chainer import utils from chainer.utils import type_check import chainerx class AbsoluteError(function_node.FunctionNode): """Element-wise absolute error function.""" def check_type_forward(self, in_types): type_check._argname(in_types, ('x0', 'x1')) type_check.expect( in_types[0].dtype.kind == 'f', in_types[0].dtype == in_types[1].dtype, in_types[0].shape == in_types[1].shape ) def forward_chainerx(self, inputs): x0, x1 = inputs self.diff = x0 - x1 return chainerx.abs(self.diff), def forward(self, inputs): x0, x1 = inputs self.diff = x0 - x1 return utils.force_array(abs(self.diff), dtype=x0.dtype), def backward(self, indexes, grad_outputs): gy, = grad_outputs gx = gy * backend.get_array_module(gy).sign(self.diff) return gx, -gx def absolute_error(x0, x1): """Element-wise absolute error function. Computes the element-wise absolute error :math:`L` between two inputs :math:`x_0` and :math:`x_1` defined as follows. .. math:: L = |x_0 - x_1| Args: x0 (:class:`~chainer.Variable` or :ref:`ndarray`): First input variable. x1 (:class:`~chainer.Variable` or :ref:`ndarray`): Second input variable. Returns: ~chainer.Variable: An array representing the element-wise absolute error between the two inputs. """ return AbsoluteError().apply((x0, x1))[0]
1,603
26.186441
77
py
chainer
chainer-master/chainer/functions/loss/sigmoid_cross_entropy.py
import chainer from chainer import backend from chainer import function_node from chainer.functions.activation import sigmoid from chainer import utils from chainer.utils import type_check class SigmoidCrossEntropy(function_node.FunctionNode): """Sigmoid activation followed by a sigmoid cross entropy loss.""" ignore_label = -1 def __init__(self, normalize=True, reduce='mean'): self.normalize = normalize if reduce not in ('mean', 'no'): raise ValueError( 'only \'mean\' and \'no\' are valid for \'reduce\', but ' '\'%s\' is given' % reduce) self.reduce = reduce self.count = None def check_type_forward(self, in_types): type_check._argname(in_types, ('x', 't')) x_type, t_type = in_types type_check.expect( x_type.dtype.kind == 'f', t_type.dtype.kind == 'i', x_type.shape == t_type.shape ) def forward(self, inputs): self.retain_inputs((0, 1)) xp = backend.get_array_module(*inputs) x, t = inputs self.ignore_mask = (t != self.ignore_label) # stable computation of the cross entropy. loss = -( self.ignore_mask * (x * (t - (x >= 0)) - xp.log1p(xp.exp(-xp.abs(x))))) if not self.reduce == 'mean': return utils.force_array(loss.astype(x.dtype)), if self.normalize: count = xp.maximum(1, self.ignore_mask.sum()) else: count = max(1, len(x)) self.count = count # TODO(takagi): Fix to perform division in a specific dtype. See # cupy/cupy#1534. return utils.force_array( xp.divide(xp.sum(loss), self.count), dtype=x.dtype), def backward(self, inputs, grad_outputs): x, t = self.get_retained_inputs() gy, = grad_outputs gx, = SigmoidCrossEntropyGrad( self.reduce, self.count, self.ignore_mask, t.data).apply((x, gy)) return gx, None class SigmoidCrossEntropyGrad(function_node.FunctionNode): """Sigmoid cross entropy gradient function.""" def __init__(self, reduce, count, ignore_mask, t): self.reduce = reduce self.count = count self.ignore_mask = ignore_mask self.t = t def forward(self, inputs): self.retain_inputs((0, 1)) xp = backend.get_array_module(*inputs) x, gy = inputs y, = sigmoid.Sigmoid().forward((x,)) if self.reduce == 'mean': # TODO(takagi): Fix to perform division in a specific dtype. See # cupy/cupy#1534. gx = xp.divide( gy * self.ignore_mask * (y - self.t), self.count).astype( y.dtype) else: gx = (gy * self.ignore_mask * (y - self.t)).astype(y.dtype) return gx, def backward(self, indexes, grad_outputs): ggx, = grad_outputs x, gy = self.get_retained_inputs() y = chainer.functions.sigmoid(x) yp = y * (1 - y) gx = yp * chainer.functions.broadcast_to(gy, yp.shape) ggy = y - self.t.astype(y.dtype) gx *= self.ignore_mask * ggx ggy *= self.ignore_mask * ggx if self.reduce == 'mean': gx /= self.count ggy = chainer.functions.sum(ggy) / self.count return gx, ggy def sigmoid_cross_entropy(x, t, normalize=True, reduce='mean'): """Computes cross entropy loss for pre-sigmoid activations. Args: x (:class:`~chainer.Variable` or :ref:`ndarray`): A variable object holding a matrix whose (i, j)-th element indicates the unnormalized log probability of the j-th unit at the i-th example. t (:class:`~chainer.Variable` or :ref:`ndarray`): A variable object holding a matrix whose (i, j)-th element indicates a signed integer vector of ground truth labels 0 or 1. If ``t[i, j] == -1``, corresponding ``x[i, j]`` is ignored. Loss is zero if all ground truth labels are ``-1``. normalize (bool): Variable holding a boolean value which determines the normalization constant. If true, this function normalizes the cross entropy loss across all instances. If else, it only normalizes along a batch size. reduce (str): Variable holding a ``str`` which determines whether to reduce the shape of the input. If it is ``'mean'``, it computes the sum of cross entropy and normalize it according to ``normalize`` option. If is is ``'no'``, this function computes cross entropy for each instance and does not normalize it (``normalize`` option is ignored). In this case, the loss value of the ignored instance, which has ``-1`` as its target value, is set to ``0``. Returns: ~chainer.Variable: A variable object holding an array of the cross entropy. If ``reduce`` is ``'mean'``, it is a scalar array. If ``reduce`` is ``'no'``, the shape is same as those of ``x`` and ``t``. .. note:: This function is differentiable only by ``x``. .. admonition:: Example >>> x = np.array([[-2.0, 3.0, 0.5], [5.0, 2.0, -0.5]]).\ astype(np.float32) >>> x array([[-2. , 3. , 0.5], [ 5. , 2. , -0.5]], dtype=float32) >>> t = np.array([[0, 1, 0], [1, 1, -1]]).astype(np.int32) >>> t array([[ 0, 1, 0], [ 1, 1, -1]], dtype=int32) >>> F.sigmoid_cross_entropy(x, t) variable(0.25664714) >>> F.sigmoid_cross_entropy(x, t, normalize=False) variable(0.64161783) >>> y = F.sigmoid_cross_entropy(x, t, reduce='no') >>> y.shape (2, 3) >>> y.array array([[ 0.126928 , 0.04858735, 0.974077 ], [ 0.00671535, 0.126928 , -0. ]], dtype=float32) """ return SigmoidCrossEntropy(normalize, reduce).apply((x, t))[0]
6,098
33.851429
77
py