Add SpeechTransformer model #374
xuzhao9
merged
38 commits into
pytorch:master
from
xuzhao9:xz9/add-speech-transformer
Add the Speech Transformer model.
79f8d79c
Adding speech transformer
e2adc2cc
WIP, need to run the model locally to get the input json.
05f9c30c
Add speech transformer files.
9c44562f
More speech_transformer fixes.
44635e0a
Fix another import issue
48a4bd8b
Fix import errors.
63d8268d
Fix cfg fields.
cada2abe
Add speech transformer input data path.
e2842104
Fix bug in dict txt load.
db4bb8d4
Change warmup steps.
346bb70c
Add another fix.
f1f0aa40
Add more arguments
2d771d0d
Fix another bug
06f03a39
Fix another bug.
a8d2700b
Fix dict path.
d8212083
Fix assert.
69963572
Enable type and disable log.
5c476f10
Disable train output.
46a0e4e8
Add real feat path
999dccf9
Add input data, remove not used files.
20a0e5f0
Fix NotImplementedError and get_modules().
439e5c7b
Return example input value
0f1001d4
Added another fix.
507745f6
Fix test bug.
4d046e90
Use cuda as the default input.
3e7e812f
Remove unused files. Add comments to explain how to generate the inpu…
fdd7927a
bitfort
approved these changes
on 2021-05-19
Use feats.1.ark as train data.
2277b40c
Preserve only feat1.
98d41c12
Fix a small bug.
0e0bf81f
Add all ark files.
32f362af
Remove solver.
778c84cc
Only preserve feats.1.ark.
640efd16
Only preserve feats.1.ark.
90009896
Fix cross validation data.
7aabfb77
Fix init function.
f70576ed
Add more notes.
e7fca8c8
Revert test.py change.
668534d6
Login to write a write a comment.
Login via GitHub