2aa4af1c29
Revert back to support input-hidden weights sharing between bi-directional RNNs.
...
1. Add options to enable and disable RNN weights sharing.
2. Set rnn_layer_size to 2048 by default.
3. Revert back the striding steps of 1st conv layer from 2 to 3.
4. Revert back to BRelu.
Above follows DS2 papers.
7 years ago
177af05953
Fix a bug in use_gru argument parsing.
7 years ago
5a63275845
Add GRU support.
7 years ago
4b3f768df7
Simplify description and codes.
7 years ago
5ef300f3f0
Make type of error rate optional.
7 years ago
6df0f9bc44
Reset default multi-thread/process number to half of cpu count() for speedup.
7 years ago
8122dd9c29
Simplify train.py, evaluate.py, infer.py and tune.py by adding DeepSpeech2Model class.
7 years ago
92eacf548b
Update default config params and result display for evaluator.py and infer.py for DS2.
7 years ago
724ef18596
update several scripts to support mfcc
7 years ago
d15c48d616
upload the language model
7 years ago
0cadc56a84
follow comments in code format
7 years ago
26510f74a6
refine ctc_beam_search_decoder
7 years ago
8033845615
add unit test for decoders
7 years ago
b2cf96cc46
adapt to the new data provider
7 years ago
ff01d048d3
final refining on old data provider: enable pruning & add evaluation & code cleanup
7 years ago
Xinghai Sun
yangyaming
Yibing Liu