[object Object][object Object][object Object]undefined[object Object]

• Description: Implements the long short-term memory (LSTM) network, which is a special recurrent neural network (RNN) model. Computes the LSTM network, receives the input sequence and initial state, and returns the output sequence and final state. Compared with , this operator adds the scenario where the sequence lengths are not equal.
• Formula:$$f_t =sigm(W_f[h_{t-1}, x_t] + b_f)\\ i_t =sigm(W_i[h_{t-1}, x_t] + b_i)\\ o_t =sigm(W_o[h_{t-1}, x_t] + b_o)\\ \tilde{c}_t =tanh(W_c[h_{t-1}, x_t] + b_c)\\ c_t =f_t ⊙ c_{t-1} + i_t ⊙ \tilde{c}_t\\ c_{o}^{t} =tanh(c_t)\\ h_t =o_t ⊙ c_{o}^{t}\\$$
  • $x_t ∈ R^{d}$: input vector to the LSTM unit.
  • $f_t ∈ (0, 1)^{h}$: activation vector of the forget gate.
  • $i_t ∈ (0, 1)^{h}$: activation vector of the input/update gate.
  • $o_t ∈ (0, 1)^{h}$: activation vector of the output gate.
  • $h_i ∈ (-1, 1)^{h}$: hidden state vector, also known as output vector of the LSTM unit.
  • $\tilde{c}_t ∈ (-1, 1)^{h}$: cell input activation vector.
  • $c_t ∈ R^{h}$: cell state vector.
  • $W ∈ R^{h×d}, (U ∈ R^{h×h})∩(b ∈ R^{h})$: weight matrices and bias vector parameters which need to be learned during training.

[object Object]

Each operator has calls. First, aclnnBidirectionLSTMV2GetWorkspaceSize is called to obtain the input parameters and compute the required workspace size based on the process. Then, aclnnBidirectionLSTMV2 is called to perform computation.

[object Object]

• Parameters:

  [object Object]

• Returns:

[object Object]: status code. For details, see .

The first-phase API implements input parameter verification. The following errors may be thrown:

[object Object][object Object]

• Parameters:

  [object Object]

• Returns:

  [object Object]: status code. For details, see .

[object Object]

• Deterministic compute:
  • aclnnBidirectionLSTMV2 defaults to a deterministic implementation.

[object Object]

The following example is for reference only. For details, see .