GRU - 3 vs 14

Next section compares an older to a newer version of the same operator after both definition are converted into markdown text. Green means an addition to the newer version, red means a deletion. Anything else is unchanged.

Files changed (1) hide show

1. GRU3 → GRU14 +7 -16

GRU3 → GRU14 RENAMED Viewed

	@@ -1 +1 @@
1 1	Computes an one-layer GRU. This operator is usually supported via some custom
2 2	implementation such as CuDNN.
3 3	Notations:
4 4	X - input tensor
5 5	z - update gate
6 6	r - reset gate
7 7	h - hidden gate
8 8	t - time step (t-1 means previous time step)
9 9	W[zrh] - W parameter weight matrix for update, reset, and hidden gates
10 10	R[zrh] - R recurrence weight matrix for update, reset, and hidden gates
11 11	Wb[zrh] - W bias vectors for update, reset, and hidden gates
12 12	Rb[zrh] - R bias vectors for update, reset, and hidden gates
13 13	WB[zrh] - W parameter weight matrix for backward update, reset, and hidden gates
14 14	RB[zrh] - R recurrence weight matrix for backward update, reset, and hidden gates
15 15	WBb[zrh] - W bias vectors for backward update, reset, and hidden gates
16 16	RBb[zrh] - R bias vectors for backward update, reset, and hidden gates
17 17	H - Hidden state
18 18	num_directions - 2 if direction == bidirectional else 1
19 19	Activation functions:
20 20	Relu(x) - max(0, x)
21 21	Tanh(x) - (1 - e^{-2x})/(1 + e^{-2x})
22 22	Sigmoid(x) - 1/(1 + e^{-x})
23 23	(NOTE: Below are optional)
24 24	Affine(x) - alpha*x + beta
25 25	LeakyRelu(x) - x if x >= 0 else alpha * x
26 26	ThresholdedRelu(x) - x if x >= alpha else 0
27 27	ScaledTanh(x) - alpha*Tanh(beta*x)
28 28	HardSigmoid(x) - min(max(alpha*x + beta, 0), 1)
29 29	Elu(x) - x if x >= 0 else alpha*(e^x - 1)
30 30	Softsign(x) - x/(1 + |x|)
31 31	Softplus(x) - log(1 + e^x)
32 32	Equations (Default: f=Sigmoid, g=Tanh):
33	- - zt = f(Xt*(Wz^T) + Ht-1*(Rz^T) + Wbz + Rbz)
33	+ - zt = f(Xt*(Wz^T) + Ht-1*Rz + Wbz + Rbz)
34	- - rt = f(Xt*(Wr^T) + Ht-1*(Rr^T) + Wbr + Rbr)
34	+ - rt = f(Xt*(Wr^T) + Ht-1*Rr + Wbr + Rbr)
35	- - ht = g(Xt*(Wh^T) + (rt (.) Ht-1)*(Rh^T) + Rbh + Wbh) # default, when linear_before_reset = 0
35	+ - ht = g(Xt*(Wh^T) + (rt (.) Ht-1)*Rh + Rbh + Wbh) # default, when linear_before_reset = 0
36	- - ht = g(Xt*(Wh^T) + (rt (.) (Ht-1*(Rh^T) + Rbh)) + Wbh) # when linear_before_reset != 0
36	+ - ht = g(Xt*(Wh^T) + (rt (.) (Ht-1*Rh + Rbh) + Wbh) # when linear_before_reset != 0
37 37	- Ht = (1 - zt) (.) ht + zt (.) Ht-1
38	- This operator has **optional** inputs/outputs. See ONNX <https://github.com/onnx/onnx/blob/master/docs/IR.md>_ for more details about the representation of optional arguments. An empty string may be used in the place of an actual argument's name to indicate a missing argument. Trailing optional arguments (those not followed by an argument that is present) may also be simply omitted.
39 38	**Attributes**
40 39	* **activation_alpha**:
41 40	Optional scaling values used by some activation functions. The
42 41	values are consumed in the order of activation functions, for
43 42	example (f, g, h) in LSTM. Default values are the same as of
44 43	corresponding ONNX operators.For example with LeakyRelu, the default
45 44	alpha is 0.01.
46 45	* **activation_beta**:
47 46	Optional scaling values used by some activation functions. The
48 47	values are consumed in the order of activation functions, for
49 48	example (f, g, h) in LSTM. Default values are the same as of
50 49	corresponding ONNX operators.
51 50	* **activations**:
52 51	A list of 2 (or 4 if bidirectional) activation functions for update,
53 52	reset, and hidden gates. The activation functions must be one of the
54 53	activation functions specified above. Optional: See the equations
55 54	for default if not specified.
56 55	* **clip**:
57 56	Cell clip threshold. Clipping bounds the elements of a tensor in the
58 57	range of [-threshold, +threshold] and is applied to the input of
59 58	activations. No clip if not specified.
60 59	* **direction**:
61 60	Specify if the RNN is forward, reverse, or bidirectional. Must be
62 61	one of forward (default), reverse, or bidirectional.
63 62	* **hidden_size**:
64 63	Number of neurons in the hidden layer
65	- * **layout**:
66	- The shape format of inputs X, initial_h and outputs Y, Y_h. If 0,
67	- the following shapes are expected: X.shape = [seq_length,
68	- batch_size, input_size], Y.shape = [seq_length, num_directions,
69	- batch_size, hidden_size], initial_h.shape = Y_h.shape =
70	- [num_directions, batch_size, hidden_size]. If 1, the following
71	- shapes are expected: X.shape = [batch_size, seq_length, input_size],
72	- Y.shape = [batch_size, seq_length, num_directions, hidden_size],
73	- initial_h.shape = Y_h.shape = [batch_size, num_directions,
74	- hidden_size].
75 64	* **linear_before_reset**:
76 65	When computing the output of the hidden gate, apply the linear
77 66	transformation before multiplying by the output of the reset gate.
67	+ * **output_sequence**:
68	+ The sequence output for the hidden is optional if 0. Default 0.
78 69	**Inputs**
79 70	Between 3 and 6 inputs.
80 71	* **X** (heterogeneous) - **T**:
81 72	The input sequences packed (and potentially padded) into one 3-D
82 73	tensor with the shape of [seq_length, batch_size, input_size].
83 74	* **W** (heterogeneous) - **T**:
84 75	The weight tensor for the gates. Concatenation of W[zrh] and
85 76	WB[zrh] (if bidirectional) along dimension 0. This tensor has
86 77	shape [num_directions, 3*hidden_size, input_size].
87 78	* **R** (heterogeneous) - **T**:
88 79	The recurrence weight tensor. Concatenation of R[zrh] and
89 80	RB[zrh] (if bidirectional) along dimension 0. This tensor has
90 81	shape [num_directions, 3*hidden_size, hidden_size].
91 82	* **B** (optional, heterogeneous) - **T**:
92 83	The bias tensor for the gates. Concatenation of [Wb[zrh], Rb[zrh]]
93 84	and [WBb[zrh], RBb[zrh]] (if bidirectional) along dimension 0.
94 85	This tensor has shape [num_directions, 6*hidden_size]. Optional:
95 86	If not specified - assumed to be 0
96 87	* **sequence_lens** (optional, heterogeneous) - **T1**:
97 88	Optional tensor specifying lengths of the sequences in a batch. If
98 89	not specified - assumed all sequences in the batch to have length
99 90	seq_length. It has shape [batch_size].
100 91	* **initial_h** (optional, heterogeneous) - **T**:
101 92	Optional initial value of the hidden. If not specified - assumed to
102 93	be 0. It has shape [num_directions, batch_size, hidden_size].
103 94	**Outputs**
104 95	Between 0 and 2 outputs.
105 96	* **Y** (optional, heterogeneous) - **T**:
106 97	A tensor that concats all the intermediate output values of the
107 98	hidden. It has shape [seq_length, num_directions, batch_size,
108	- hidden_size].
99	+ hidden_size]. It is optional if output_sequence is 0.
109 100	* **Y_h** (optional, heterogeneous) - **T**:
110 101	The last output value of the hidden. It has shape [num_directions,
111 102	batch_size, hidden_size].
112 103	**Type Constraints**
113 104	* **T** in (
114 105	tensor(double),
115 106	tensor(float),
116 107	tensor(float16)
117 108	):
118 109	Constrain input and output types to float tensors.
119 110	* **T1** in (
120 111	tensor(int32)
121 112	):
122 113	Constrain seq_lens to integer tensor.