File: gru_unit_op.cc

package info (click to toggle)
pytorch 1.13.1%2Bdfsg-4
  • links: PTS, VCS
  • area: main
  • in suites: bookworm
  • size: 139,252 kB
  • sloc: cpp: 1,100,274; python: 706,454; ansic: 83,052; asm: 7,618; java: 3,273; sh: 2,841; javascript: 612; makefile: 323; xml: 269; ruby: 185; yacc: 144; objc: 68; lex: 44
file content (55 lines) | stat: -rw-r--r-- 1,894 bytes parent folder | download | duplicates (2)
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
#include "gru_unit_op.h"

namespace caffe2 {
REGISTER_CPU_OPERATOR(GRUUnit, GRUUnitOp<float, CPUContext>);
OPERATOR_SCHEMA(GRUUnit)
    .NumInputs(3, 4)
    .NumOutputs(1)
    .SetDoc(R"DOC(
GRUUnit computes the activations of a standard GRU,
in a sequence-length aware fashion.

Concretely, given the (fused) inputs X (TxNxD), the previous hidden
state (NxD), and the sequence lengths (N), computes the GRU
activations, avoiding computation if the input is invalid (as in, the
value at X[t][n] >= seqLengths[n].

)DOC")
    .Arg(
        "drop_states",
        "Bool to determine if hidden state is zeroes or passed "
        "along for timesteps past the given sequence_length.")
    .Arg(
        "sequence_lengths",
        "When false, the sequence lengths input is left out, "
        "and all following inputs are shifted left by one.")
    .Output(0, "hidden", "The new GRU hidden state calculated by this op.");
REGISTER_CPU_OPERATOR(GRUUnitGradient, GRUUnitGradientOp<float, CPUContext>);
OPERATOR_SCHEMA(GRUUnitGradient)
    .NumInputs(5, 6)
    .NumOutputs(2)
    .Arg(
        "sequence_lengths",
        "When false, the sequence lengths input is left out, "
        "and all following inputs are shifted left by one.");

class GetGRUUnitGradient : public GradientMakerBase {
  using GradientMakerBase::GradientMakerBase;
  vector<OperatorDef> GetGradientDefs() override {
    if (GetFlagArgument(def_, "sequence_lengths", true)) {
      return SingleGradientDef(
          "GRUUnitGradient",
          "",
          vector<string>{I(0), I(1), I(2), I(3), O(0), GO(0)},
          vector<string>{GI(0), GI(1)});
    } else {
      return SingleGradientDef(
          "GRUUnitGradient",
          "",
          vector<string>{I(0), I(1), I(2), O(0), GO(0)},
          vector<string>{GI(0), GI(1)});
    }
  }
};
REGISTER_GRADIENT(GRUUnit, GetGRUUnitGradient);
} // namespace caffe2