Class: TensorStream::MathGradients

Inherits:
Object
  • Object
show all
Extended by:
OpHelper
Defined in:
lib/tensor_stream/math_gradients.rb

Overview

Class that provides auto-differentiation Most gradients are ported over from tensorflow’s math_grad.py

Class Method Summary collapse

Methods included from OpHelper

_op, cons, format_source, fp_type?, i_cons, i_op, int_type?, reduced_shape, shape_eval, shape_full_specified, shapes_fully_specified_and_equal

Class Method Details

._broadcast_gradient_args(input_a, input_b) ⇒ Object 



330
331
332
333
 
# File 'lib/tensor_stream/math_gradients.rb', line 330

def self._broadcast_gradient_args(input_a, input_b)
  res = _op(:broadcast_gradient_args, input_a, input_b)
  [res[0], res[1]]
end

._broadcast_mul(vec, mat) ⇒ Object 



376
377
378
379
 
# File 'lib/tensor_stream/math_gradients.rb', line 376

def self._broadcast_mul(vec, mat)
  vec = ts.expand_dims(vec, -1)
  vec * mat
end

._broadcast_transform(input_a, input_b) ⇒ Object 



335
336
337
 
# File 'lib/tensor_stream/math_gradients.rb', line 335

def self._broadcast_transform(input_a, input_b)
  _op(:broadcast_transform, input_a, input_b)
end

._compute_derivative(node, grad) ⇒ Object

TODO: refactor and implement registerGradient



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
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
 
# File 'lib/tensor_stream/math_gradients.rb', line 49

def self._compute_derivative(node, grad)
  node.graph.name_scope("#{node.name}_grad") do
    x = node.inputs[0] if node.inputs[0]
    y = node.inputs[1] if node.inputs[1]

    case node.operation
    when :add_n
      return [grad] * node.inputs.size
    when :add
      return [grad, grad] if shapes_fully_specified_and_equal(x, y)
      sx = ts.shape(x, name: 'add/shape_x')
      sy = ts.shape(y, name: 'add/shape_y')
      rx, ry = _broadcast_gradient_args(sx, sy)

      [ts.reshape(ts.reduce_sum(grad, rx, name: 'add/reduce_sum_x'), sx),
       ts.reshape(ts.reduce_sum(grad, ry, name: 'add/reduce_sum_y'), sy)]
    when :asin
      ts.control_dependencies([grad]) do
        x2 = ts.square(x)
        one = ts.constant(1, dtype: grad.data_type)
        den = ts.sqrt(ts.subtract(one, x2))
        inv = ts.reciprocal(den)
        grad * inv
      end
    when :acos
      ts.control_dependencies([grad]) do
        x2 = ts.square(x)
        one = ts.constant(1, dtype: grad.data_type)
        den = ts.sqrt(ts.subtract(one, x2))
        inv = ts.reciprocal(den)
        -grad * inv
      end
    when :atan
      ts.control_dependencies([grad]) do
        x2 = ts.square(x)
        one = ts.constant(1, dtype: grad.data_type)
        inv = ts.reciprocal(ts.add(one, x2))
        grad * inv
      end
    when :fill
      [nil, ts.reduce_sum(grad)]
    when :sub
      return [grad, -grad] if shapes_fully_specified_and_equal(x, y)

      sx = ts.shape(x, name: 'sub/shape_x')
      sy = ts.shape(y, name: 'sub/shape_y')
      rx, ry = _broadcast_gradient_args(sx, sy)

      [ts.reshape(ts.reduce_sum(grad, rx, name: 'add/reduce_sub_x'), sx),
       -ts.reshape(ts.reduce_sum(grad, ry, name: 'add/reduce_sub_y'), sy)]
    when :mul
      sx = ts.shape(x)
      sy = ts.shape(y)
      rx, ry = _broadcast_gradient_args(sx, sy)

      [ts.reshape(ts.reduce_sum(ts.mul(grad, y), rx), sx),
       ts.reshape(ts.reduce_sum(ts.mul(x, grad), ry), sy)]
    when :div
      sx = i_op(:shape, x)
      sy = i_op(:shape, y)
      rx, ry = _broadcast_gradient_args(sx, sy)

      [ts.reshape(ts.reduce_sum(ts.div(grad, y), rx), sx),
       ts.reshape(ts.reduce_sum(grad * ts.div(ts.div(-x, y), y), ry), sy)]
    when :mod
      sx = ts.shape(x)
      sy = ts.shape(y)
      rx, ry = _broadcast_gradient_args(sx, sy)
      floor_xy = ts.floor_div(x, y)
      gx = ts.reshape(ts.reduce_sum(grad, rx), sx)
      gy = ts.reshape(ts.reduce_sum(grad * ts.negative(floor_xy), ry), sy)

      [gx, gy]
    when :prod
      input_shape = ts.shape(x)
      y = ts.range(0, ts.rank(x)) if y.nil?
      reduction_indices = ts.reshape(y, [-1])

      output_shape_kept_dims = ts.reduced_shape(input_shape, y)
      tile_scaling = _safe_shape_div(input_shape, output_shape_kept_dims)
      grad = ts.reshape(grad, output_shape_kept_dims)
      grad = ts.tile(grad, tile_scaling)

      perm, reduced_num, other_num = ts.device("/cpu:0") do
        rank = ts.rank(x)
        reduction_indices = (reduction_indices + rank) % rank
        reduced = ts.cast(reduction_indices, :int32)
        idx = ts.range(0, rank)
        other, = ts.setdiff1d(idx, reduced)

        [ts.concat([reduced, other], 0),
         ts.reduce_prod(ts.gather(input_shape, reduced)),
         ts.reduce_prod(ts.gather(input_shape, other))]
      end

      permuted = ts.transpose(x, perm)
      permuted_shape = ts.shape(permuted)

      reshaped = ts.reshape(permuted, [reduced_num, other_num])

      # Calculate product, leaving out the current entry
      left = ts.cumprod(reshaped, axis: 0, exclusive: true)
      right = ts.cumprod(reshaped, axis: 0, exclusive: true, reverse: true)
      y = ts.reshape(left * right, permuted_shape)

      # Invert the transpose and reshape operations.
      # Make sure to set the statically known shape information through a reshape.
      out = grad * ts.transpose(y, ts.invert_permutation(perm))
      [ts.reshape(out, input_shape, name: 'prod'), nil]
    when :squared_difference
      sx = i_op(:shape, x)
      sy = i_op(:shape, y)
      rx, ry = _broadcast_gradient_args(sx, sy)

      x_grad = ts.mul(2.0, grad) * (x - y)

      [ts.reshape(ts.reduce_sum(x_grad, rx), sx),
       ts.reshape(-ts.reduce_sum(x_grad, ry), sy)]
    when :mat_mul
      t_a = node.options[:transpose_a]
      t_b = node.options[:transpose_b]

      if !t_a && !t_b
        grad_a = ts.matmul(grad, y, transpose_b: true)
        grad_b = ts.matmul(x, grad, transpose_a: true)
      elsif !ta && tb
        grad_a = ts.matmul(grad, y)
        grad_b = ts.matmul(grad, x, transpose_a: true)
      elsif t_a && !t_b
        grad_a = ts.matmul(y, grad, transpose_b: true)
        grad_b = ts.matmul(x, grad)
      elsif t_a && t_b
        grad_a = ts.matmul(y, grad, transpose_a: true, transpose_b: true)
        grad_b = ts.matmul(grad, x, transpose_a: true, transpose_b: true)
      end

      [grad_a, grad_b]
    when :sin
      grad * ts.cos(x)
    when :tanh
      grad * i_op(:tanh_grad, x)
    when :pow
      z = node
      sx = ts.shape(x)
      sy = ts.shape(y)
      rx, ry = _broadcast_gradient_args(sx, sy)
      gx = ts.reduce_sum(grad * y * ts.pow(x, y - 1), rx)

      log_x = ts.where(x > 0, ts.log(x), ts.zeros_like(x))
      gy = ts.reduce_sum(grad * z * log_x, ry)

      [gx, gy]
    when :abs
      grad * ts.sign(x)
    when :log
      grad * ts.reciprocal(x)
    when :cos
      -grad * ts.sin(x)
    when :max
      _min_or_max_grad(node.inputs, grad, ->(a, b) { ts.greater_equal(a, b) })
    when :min
      _min_or_max_grad(node.inputs, grad, ->(a, b) { ts.less_equal(a, b) })
    when :tan
      secx = ts.reciprocal(ts.cos(x))
      secx2 = ts.square(secx)
      grad * secx2
    when :negate
      -grad
    when :exp
      grad * node
    when :identity, :print
      grad
    when :sign
      ts.zeros(ts.shape(x), dtype: x.data_type)
    when :tile
      input_shape = ts.shape(x)
      split_shape = ts.reshape(ts.transpose(ts.stack([y, input_shape])), [-1])
      axes = ts.range(0, ts.size(split_shape), 2)
      input_grad = ts.reduce_sum(ts.reshape(grad, split_shape), axes)

      [input_grad, nil]
    when :sum
      _sum_grad(x, y, grad)
    when :reciprocal
      -grad * (ts.constant(1, dtype: x.dtype) / x**2)
    when :sqrt
      ts.constant(1, dtype: x.dtype) / (ts.constant(2, dtype: x.dtype) * ts.sqrt(x)) * grad
    when :stop_gradient
      ts.zeros_like(grad)
    when :square
      y = ts.constant(2.0, dtype: x.dtype)
      ts.multiply(grad, ts.multiply(x, y))
    when :where
      x_mask = i_op(:where, i_op(:ones_like, x), i_op(:zeros_like, y), pred: node.options[:pred])
      y_mask = i_op(:where, i_op(:zeros_like, x), i_op(:ones_like, y), pred: node.options[:pred])
      [x_mask * grad, y_mask * grad]
    when :cond
      x_cond = i_op(:cond, i_op(:ones_like, x), i_op(:zeros_like, y), pred: node.options[:pred])
      y_cond = i_op(:cond, i_op(:zeros_like, x), i_op(:ones_like, x), pred: node.options[:pred])
      [x_cond * grad, y_cond * grad]
    when :mean
      sum_grad = _sum_grad(x, y, grad)[0]
      input_shape = ts.shape(x)
      output_shape = ts.shape(node)
      factor = _safe_shape_div(ts.reduce_prod(input_shape), ts.reduce_prod(output_shape))
      [ts.div(sum_grad, ts.cast(factor, sum_grad.data_type)), nil]
    when :log1p
      grad * ts.reciprocal(i_cons(1, dtype: grad.data_type) + x)
    when :sigmoid
      i_op(:sigmoid_grad, x, grad)
    when :sigmoid_grad
      gb = grad * y
      [gb - 2.0 * gb * x, i_op(:sigmoid_grad, x, grad)]
    when :softmax
      i_op(:softmax_grad, x, grad)
    when :softmax_cross_entropy_with_logits_v2
      output = node
      logits = node.inputs[0]
      [_broadcast_mul(grad, output[1]), -ts.nn.log_softmax(logits)]
    when :sparse_softmax_cross_entropy_with_logits
      output = node
      [_broadcast_mul(grad, output[1]), nil]
    when :floor, :ceil, :round
      # non differentiable
      nil
    when :zeros_like
      # non differentiable
      nil
    when :argmin, :argmax, :floor_div
      # non differentiable
      [nil, nil]
    when :transpose
      return [ts.transpose(grad, ts.invert_permutation(y)), nil]
    when :index
      # hack!! not sure how to fix this yet
      return grad if %i[softmax_cross_entropy_with_logits_v2 sparse_softmax_cross_entropy_with_logits].include?(node.inputs[0].operation)

      if node.inputs[0].shape.known? && node.inputs[1].value
        multiplier = node.inputs[0].shape.shape[0]
        filler = ts.zeros_like(grad)

        res = Array.new(multiplier) do |index|
          index == node.inputs[1].value ? grad : filler
        end
        [res]
      end
    when :squeeze
      _reshape_to_input(node, grad)
    when :expand_dims
      [_reshape_to_input(node, grad), nil]
    when :concat
      _concat_grad_helper(node, grad, 1, node.inputs.size, 0)
    when :reshape
      [ts.reshape(grad, ts.shape(node.inputs[0])), nil]
    when :stack
      res = ts.unstack(grad, num: node.inputs.size, axis: node.options[:axis])
      Array.new(node.inputs.size) { |i| res[i] }
    when :unstack
      ts.stack(grad, axis: node.options[:axis])
    when :conv2d
      _Conv2DGrad(node, grad)
    when :cast
      t = %i[float16 float32 float64]
      src_type = node.inputs[0].data_type
      dst_type = grad.data_type

      if t.key?(src_type) && t.key?(dst_type)
        ts.cast(grad, src_type)
      end

      nil
    else
      raise "no derivative op for #{node.operation}"
    end
  end
end

._concat_grad_helper(op, grad, start_value_index, end_value_index, dim_index) ⇒ Object 



405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
 
# File 'lib/tensor_stream/math_gradients.rb', line 405

def self._concat_grad_helper(op, grad, start_value_index, end_value_index, dim_index)
  # Degenerate concatenation, just return grad.
  if op.inputs.size == 2
    return end_value_index <= dim_index ? [grad] + [nil] : [nil] + [grad]
  end
  concat_dim = op.inputs[dim_index]
  input_values = op.inputs[start_value_index..end_value_index]
  non_neg_concat_dim = concat_dim % ts.rank(input_values[0])
  sizes = _extract_input_shapes(input_values)

  slicer = ts.slice(ts.stack(sizes, axis: 1), [non_neg_concat_dim, 0], [1, -1])
  sizes = ts.squeeze(slicer)

  out_grads = ts.split(grad, sizes, axis: non_neg_concat_dim, num: op.inputs.size - 1)
  end_value_index <= dim_index ? out_grads + [nil] : [nil] + out_grads
end

._Conv2DGrad(op, grad) ⇒ Object 



422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
 
# File 'lib/tensor_stream/math_gradients.rb', line 422

def self._Conv2DGrad(op, grad)
  # dilations = op.get_attr("dilations")
  strides = op.options[:strides]
  padding = op.options[:padding]
  use_cudnn_on_gpu = op.options[:use_cudnn_on_gpu]
  data_format = op.options[:data_format]

  shape_0, shape_1 = ts.shape_n([op.inputs[0], op.inputs[1]])
  [
      _op(:conv2d_backprop_input,
          shape_0,
          op.inputs[1],
          grad,
          strides: strides,
          padding: padding,
          use_cudnn_on_gpu: use_cudnn_on_gpu,
          data_format: data_format),
      _op(:conv2d_backprop_filter,
          op.inputs[0],
          shape_1,
          grad,
          strides: strides,
          padding: padding,
          use_cudnn_on_gpu: use_cudnn_on_gpu,
          data_format: data_format)
  ]
end

._extract_input_shapes(inputs) ⇒ Object 



386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
 
# File 'lib/tensor_stream/math_gradients.rb', line 386

def self._extract_input_shapes(inputs)
  sizes = []
  fully_known = true
  inputs.each do |x|
    input_shape = ts.shape(x)
    unless input_shape.is_const
      fully_known = false
      break
    end
    sizes << input_shape.value
  end

  if fully_known
    sizes
  else
    ts.shape_n(inputs)
  end
end

._include?(arr, obj) ⇒ Boolean

Returns:



381
382
383
384
 
# File 'lib/tensor_stream/math_gradients.rb', line 381

def self._include?(arr, obj)
  arr.each { |a| return true if a.equal?(obj) }
  false
end

._min_or_max_grad(inputs, grad, selector_op) ⇒ Object 



359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
 
# File 'lib/tensor_stream/math_gradients.rb', line 359

def self._min_or_max_grad(inputs, grad, selector_op)
  x = inputs[0]
  y = inputs[1]
  gdtype = grad.data_type
  sx = ts.shape(x)
  sy = ts.shape(y)
  gradshape = ts.shape(grad)
  zeros = ts.zeros(gradshape, dtype: gdtype)
  xmask = selector_op.call(x, y)
  rx, ry = _broadcast_gradient_args(sx, sy)
  xgrad = ts.where(xmask, grad, zeros, name: 'x')
  ygrad = ts.where(xmask, zeros, grad, name: 'y')
  gx = ts.reshape(ts.reduce_sum(xgrad, rx), sx)
  gy = ts.reshape(ts.reduce_sum(ygrad, ry), sy)
  [gx, gy]
end

._op_supports_broadcast?(node) ⇒ Boolean

Returns:



354
355
356
357
 
# File 'lib/tensor_stream/math_gradients.rb', line 354

def self._op_supports_broadcast?(node)
  return true if %i[add sub div mul pow].include?(node.operation)
  false
end

._propagate(grad, tensor, stop_tensor, nodes_to_compute, stop_gradients = []) ⇒ Object 



25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
 
# File 'lib/tensor_stream/math_gradients.rb', line 25

def self._propagate(grad, tensor, stop_tensor, nodes_to_compute, stop_gradients = [])
  return grad if stop_tensor.equal?(tensor)
  return nil if stop_gradients && _include?(stop_gradients, tensor)
  return nil unless tensor.is_a?(Operation)

  computed_op = _compute_derivative(tensor, grad)

  if computed_op.is_a?(Array)
    grads = computed_op.each_with_index.collect do |op_grad, index|
      next if op_grad.nil?
      next unless nodes_to_compute.include?(tensor.inputs[index].name)

      _propagate(op_grad, tensor.inputs[index], stop_tensor, nodes_to_compute, stop_gradients)
    end.compact

    return nil if grads.empty?
    grads.size > 1 ? ts.add_n(grads) : grads[0]
  else
    return nil if computed_op.nil?
    _propagate(computed_op, tensor.inputs[0], stop_tensor, nodes_to_compute, stop_gradients)
  end
end

._reshape_to_input(node, grad) ⇒ Object 



326
327
328
 
# File 'lib/tensor_stream/math_gradients.rb', line 326

def self._reshape_to_input(node, grad)
  ts.reshape(grad, ts.shape(node.inputs[0]))
end

._safe_shape_div(arg_x, arg_y) ⇒ Object 



339
340
341
 
# File 'lib/tensor_stream/math_gradients.rb', line 339

def self._safe_shape_div(arg_x, arg_y)
  _op(:floor_div, arg_x, ts.maximum(arg_y, 1))
end

._sum_grad(arg_x, arg_y, grad) ⇒ Object 



343
344
345
346
347
348
349
350
351
352
 
# File 'lib/tensor_stream/math_gradients.rb', line 343

def self._sum_grad(arg_x, arg_y, grad)
  input_shape = _op(:shape, arg_x)
  output_shape_kept_dims = ts.reduced_shape(input_shape, arg_y)
  tile_scaling = _safe_shape_div(input_shape, output_shape_kept_dims)
  new_grad = _op(:reshape, grad, output_shape_kept_dims)

  grad = _op(:cond, _op(:fill, input_shape, grad), _op(:tile, new_grad, tile_scaling), pred: _op(:rank, grad).zero?)

  [grad, nil]
end

.derivative(tensor, wrt_dx, options = {}) ⇒ Object 



11
12
13
14
15
16
17
18
19
20
21
22
23
 
# File 'lib/tensor_stream/math_gradients.rb', line 11

def self.derivative(tensor, wrt_dx, options = {})
  return i_op(:ones_like, tensor) if tensor.equal?(wrt_dx)
  return i_op(:zeros_like, wrt_dx) unless wrt_dx.consumers.include?(tensor.name)

  nodes_to_compute = wrt_dx.consumers.select do |t|
    node = tensor.graph.nodes[t]
    node.consumers.include?(tensor.name) || node.equal?(tensor)
  end.compact + [wrt_dx.name]

  grad = i_op(:fill, ts.shape(tensor), ts.constant(1, dtype: wrt_dx.data_type))

  _propagate(grad, tensor, wrt_dx, nodes_to_compute, options[:stop_gradients] || []) || i_op(:zeros_like, wrt_dx)
end

.ts ⇒ Object 



7
8
9
 
# File 'lib/tensor_stream/math_gradients.rb', line 7

def self.ts
  TensorStream
end