RNN - 1 vs 7

Files changed (1) hide show

1. RNN1 → RNN7 +3 -4

RNN1 → RNN7 RENAMED Viewed

	@@ -1 +1 @@
1 1	Computes an one-layer simple RNN. This operator is usually supported
2 2	via some custom implementation such as CuDNN.
3 3	Notations:
4 4	X - input tensor
5 5	i - input gate
6 6	t - time step (t-1 means previous time step)
7 7	Wi - W parameter weight matrix for input gate
8 8	Ri - R recurrence weight matrix for input gate
9 9	Wbi - W parameter bias vector for input gate
10 10	Rbi - R parameter bias vector for input gate
11 11	WBi - W parameter weight matrix for backward input gate
12 12	RBi - R recurrence weight matrix for backward input gate
13 13	WBbi - WR bias vectors for backward input gate
14 14	RBbi - RR bias vectors for backward input gate
15 15	H - Hidden state
16 16	num_directions - 2 if direction == bidirectional else 1
17 17	Activation functions:
18 18	Relu(x) - max(0, x)
19 19	Tanh(x) - (1 - e^{-2x})/(1 + e^{-2x})
20 20	Sigmoid(x) - 1/(1 + e^{-x})
21 21	(NOTE: Below are optional)
22 22	Affine(x) - alpha*x + beta
23 23	LeakyRelu(x) - x if x >= 0 else alpha * x
24 24	ThresholdedRelu(x) - x if x >= alpha else 0
25 25	ScaledTanh(x) - alpha*Tanh(beta*x)
26 26	HardSigmoid(x) - min(max(alpha*x + beta, 0), 1)
27 27	Elu(x) - x if x >= 0 else alpha*(e^x - 1)
28 28	Softsign(x) - x/(1 + |x|)
29 29	Softplus(x) - log(1 + e^x)
30 30	Equations (Default: f=Tanh):
31	- - Ht = f(Xt*(Wi^T) + Ht-1*Ri + Wbi + Rbi)
31	+ - Ht = f(Xt*(Wi^T) + Ht-1*(Ri^T) + Wbi + Rbi)
32	+ 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.
32 33	**Attributes**
33 34	* **activation_alpha**:
34 35	Optional scaling values used by some activation functions. The
35 36	values are consumed in the order of activation functions, for
36 37	example (f, g, h) in LSTM. Default values are the same as of
37 38	corresponding ONNX operators.For example with LeakyRelu, the default
38 39	alpha is 0.01.
39 40	* **activation_beta**:
40 41	Optional scaling values used by some activation functions. The
41 42	values are consumed in the order of activation functions, for
42 43	example (f, g, h) in LSTM. Default values are the same as of
43 44	corresponding ONNX operators.
44 45	* **activations**:
45 46	One (or two if bidirectional) activation function for input gate.
46 47	The activation function must be one of the activation functions
47 48	specified above. Optional: Default Tanh if not specified.
48 49	* **clip**:
49 50	Cell clip threshold. Clipping bounds the elements of a tensor in the
50 51	range of [-threshold, +threshold] and is applied to the input of
51 52	activations. No clip if not specified.
52 53	* **direction**:
53 54	Specify if the RNN is forward, reverse, or bidirectional. Must be
54 55	one of forward (default), reverse, or bidirectional.
55 56	* **hidden_size**:
56 57	Number of neurons in the hidden layer
57	- * **output_sequence**:
58	- The sequence output for the hidden is optional if 0. Default 0.
59 58	**Inputs**
60 59	Between 3 and 6 inputs.
61 60	* **X** (heterogeneous) - **T**:
62 61	The input sequences packed (and potentially padded) into one 3-D
63 62	tensor with the shape of [seq_length, batch_size, input_size].
64 63	* **W** (heterogeneous) - **T**:
65 64	The weight tensor for input gate. Concatenation of Wi and WBi
66 65	(if bidirectional). The tensor has shape [num_directions,
67 66	hidden_size, input_size].
68 67	* **R** (heterogeneous) - **T**:
69 68	The recurrence weight tensor. Concatenation of Ri and RBi (if
70 69	bidirectional). The tensor has shape [num_directions, hidden_size,
71 70	hidden_size].
72 71	* **B** (optional, heterogeneous) - **T**:
73 72	The bias tensor for input gate. Concatenation of [Wbi, Rbi] and
74 73	[WBbi, RBbi] (if bidirectional). The tensor has shape
75 74	[num_directions, 2*hidden_size]. Optional: If not specified -
76 75	assumed to be 0.
77 76	* **sequence_lens** (optional, heterogeneous) - **T1**:
78 77	Optional tensor specifying lengths of the sequences in a batch. If
79 78	not specified - assumed all sequences in the batch to have length
80 79	seq_length. It has shape [batch_size].
81 80	* **initial_h** (optional, heterogeneous) - **T**:
82 81	Optional initial value of the hidden. If not specified - assumed to
83 82	be 0. It has shape [num_directions, batch_size, hidden_size].
84 83	**Outputs**
85 84	Between 0 and 2 outputs.
86 85	* **Y** (optional, heterogeneous) - **T**:
87 86	A tensor that concats all the intermediate output values of the
88 87	hidden. It has shape [seq_length, num_directions, batch_size,
89	- hidden_size]. It is optional if output_sequence is 0.
88	+ hidden_size].
90 89	* **Y_h** (optional, heterogeneous) - **T**:
91 90	The last output value of the hidden. It has shape [num_directions,
92 91	batch_size, hidden_size].
93 92	**Type Constraints**
94 93	* **T** in (
95 94	tensor(double),
96 95	tensor(float),
97 96	tensor(float16)
98 97	):
99 98	Constrain input and output types to float tensors.
100 99	* **T1** in (
101 100	tensor(int32)
102 101	):
103 102	Constrain seq_lens to integer tensor.