MCPcopy Create free account

hub / github.com/HawkAaron/RNN-Transducer / functions

Functions82 in github.com/HawkAaron/RNN-Transducer

⨍ Functions 82◇ Types & classes 17
↓ 6 callersMethodbackward
(self, req, out_grad, in_data, out_data, in_grad, aux)
rnnt_np.py:110
↓ 6 callersFunctionlogsumexp
Stable log sum exp.
ctc_decoder.py:27
↓ 5 callersMethodjoint
`f`: encoder lstm output (B,T,U,2H) expanded `g`: decoder lstm output (B,T,U,H) expanded NOTE f and g must have the same size except
model.py:56
↓ 3 callersMethoddecode
(self, xs, hidden)
model2012.py:21
↓ 2 callersMethod_dump
(self)
DataLoader.py:59
↓ 2 callersFunctionconvert
(inputs, labels, ctx)
DataLoader.py:30
↓ 2 callersFunctionremap
(y, blank)
eval.py:51
↓ 1 callersMethod__init__
(self, vocab_size, num_hidden, num_layers, dropout=0, bidirectional=False)
model2012.py:10
↓ 1 callersMethod__init__
(self, blank=0, weight=None, **kwargs)
rnnt_np.py:134
↓ 1 callersMethod__init__
(self, vocab_size, num_hidden, num_layers, dropout=0, bidirectional=False)
model.py:11
↓ 1 callersMethod_distance
(self, y, t)
DataLoader.py:106
↓ 1 callersMethod_step
(self, inputs, hidden, enc_out)
seq2seq/decoder.py:47
↓ 1 callersFunctionbackward_pass
(log_probs, labels, blank)
rnnt_np.py:23
↓ 1 callersMethodbeam_search
(self, xs, W)
model.py:33
↓ 1 callersFunctioncompute_gradient
(log_probs, alphas, betas, labels, blank)
rnnt_np.py:43
↓ 1 callersFunctiondecode
()
eval.py:62
↓ 1 callersFunctiondistance
(y, t, blank=rephone[0])
eval.py:50
↓ 1 callersFunctionend_pad_concat
(inputs)
DataLoader.py:22
↓ 1 callersFunctionevaluate
(ctx=context)
train_att.py:71
↓ 1 callersFunctionevaluate
(ctx=context)
train.py:80
↓ 1 callersFunctionevaluate
(ctx=context)
train_ctc.py:87
↓ 1 callersFunctionforward_pass
(log_probs, labels, blank)
rnnt_np.py:4
↓ 1 callersMethodgreedy_decode
(self, xs)
model.py:26
↓ 1 callersFunctionlog_aplusb
(a, b)
model2012.py:158
↓ 1 callersFunctionlog_aplusb
(a, b)
model.py:185
↓ 1 callersFunctionloss_test
()
rnnt_np.py:174
↓ 1 callersFunctionmake_new_beam
()
ctc_decoder.py:23
↓ 1 callersFunctiontrain
()
train_att.py:78
↓ 1 callersFunctiontrain
()
train.py:87
↓ 1 callersFunctiontrain
()
train_ctc.py:95
↓ 1 callersFunctiontransduce
Args: log_probs: 3D array with shape [input len, output len + 1, vocab size] labels: 1D array with shape [output ti
rnnt_np.py:57
↓ 1 callersFunctiontransduce_batch
(log_probs, labels, flen, glen, blank=0)
rnnt_np.py:73
↓ 1 callersFunctionzero_pad_concat
(inputs)
DataLoader.py:14
Method__init__
(self, batch_first=True, blank_label=0, weight=None, **kwargs)
rnnt_mx.py:5
Method__init__
(self, vocab_size, num_hidden, num_layers, dropout=0, blank=0, bidirectional=False)
model2012.py:27
Method__init__
(self, seq=None, hidden=None, blank=0)
model2012.py:163
Method__init__
(self, layout='NTC', label_layout='NT', weight=None, **kwargs)
train_ctc.py:14
Method__init__
(self, dtype, batch_size=1, ctx=mx.cpu(0), attention=False)
DataLoader.py:39
Method__init__
(self, blank=0)
DataLoader.py:79
Method__init__
(self, blank)
rnnt_np.py:93
Method__init__
(self, blank=0)
rnnt_np.py:117
Method__init__
(self, vocab_size, num_hidden, num_layers, dropout=0, blank=0, bidirectional=False)
model.py:40
Method__init__
(self, seq=None, hidden=None, blank=0)
model.py:190
Method__init__
(self, vocab_size, hidden_size, num_layers, sample_rate, **kwargs)
seq2seq/decoder.py:8
Method__init__
(self, vocab_size, hidden_size, num_layers, dropout, sample_rate=0.4, **kwargs)
seq2seq/seq2seq.py:11
Method__init__
(self, **kwargs)
seq2seq/attention.py:5
Method__init__
(self, hidden_size, num_layers, dropout, **kwargs)
seq2seq/encoder.py:6
Method__iter__
(self)
DataLoader.py:68
Method__len__
(self)
DataLoader.py:56
Method__str__
(self)
model2012.py:176
Method__str__
(self)
model.py:203
Methodbeam_search
`xs`: acoustic model outputs NOTE only support one sequence (batch size = 1)
model2012.py:76
Methodbeam_search
`xs`: acoustic model outputs NOTE only support one sequence (batch size = 1)
model.py:100
Methodcreate_operator
(self, ctx, shapes, dtypes)
rnnt_np.py:130
Functiondecode
Performs inference for the given output probabilities. Arguments: probs: The output probabilities (e.g. log post-softmax) for each
ctc_decoder.py:38
Methoddecode
(self, xs, hidden)
model.py:22
Methodforward
(self, xs)
model2012.py:17
Methodforward
(self, xs, ys, xlen, ylen)
model2012.py:40
Methodforward
`log_ytu`: am & pm joint probability, layout 'BTUV' `y`: label sequence (blank, y1, ..., yU), layout 'BU' `flen`: acoustic mo
rnnt_np.py:96
Methodforward
(self, xs)
model.py:18
Methodforward
(self, xs, ys, xlen, ylen)
model.py:65
Methodforward
`target`: (batch, length) `enc_out`: Encoder output, (batch, length, hidden_size) `enc_hid`: last hidden state of encoder
seq2seq/decoder.py:22
Methodforward
`inputs`: (batch, length, dim) `targets`: (batch, length)
seq2seq/seq2seq.py:19
Methodforward
(self, hidden, enc_out)
seq2seq/attention.py:8
Methodforward
`inputs`: (batch, length, input_size) `hidden`: Initial hidden state (num_layer, batch_size, hidden_size)
seq2seq/encoder.py:13
Methodforward_step
`label`: int
model2012.py:81
Methodforward_step
`label`: int
model.py:106
Methodget
(self, err=True)
DataLoader.py:95
MethodgetAll
(self, err=True)
DataLoader.py:102
Methodgreedy_decode
TODO batch support / gpu support `weight`: acoustic score weight
model2012.py:54
Methodgreedy_decode
(self, xs)
model.py:81
Methodhybrid_forward
(self, F, pred, label, pred_lengths, label_lengths)
rnnt_mx.py:11
Methodhybrid_forward
(self, F, pred, label, pred_lengths=None, label_lengths=None)
train_ctc.py:24
Methodhybrid_forward
(self, F, log_ytu, label, flen, glen)
rnnt_np.py:139
Methodinfer_shape
(self, in_shapes)
rnnt_np.py:127
Methodisprefix
(a, b)
model2012.py:87
Methodisprefix
(a, b)
model.py:112
Functionjoint_test
()
rnnt_np.py:146
Methodlist_arguments
(self)
rnnt_np.py:121
Methodlist_outputs
(self)
rnnt_np.py:124
Functionseperate_test
()
rnnt_np.py:157
Methodupdate
(self, pred, label, xlen, ylen)
DataLoader.py:86