Daniel Povey
07d3369234
Finish feat-diagonalizing code
2022-05-17 14:13:56 +08:00
Daniel Povey
2f2934a115
fix train.py
2022-05-16 19:41:16 +08:00
Daniel Povey
d61c8aa3bc
Hopefully this finishes the full orthogonalization.
2022-05-16 19:18:12 +08:00
Daniel Povey
67f916e599
Draft towards 2nd orthogonalization
2022-05-16 16:16:12 +08:00
Daniel Povey
8aeaf1421a
Rename orthogonalize to diagonalize
2022-05-16 12:43:47 +08:00
Daniel Povey
9859e33c06
Remove optimizer.reset() which is not supported for Eve
2022-05-16 12:13:36 +08:00
Daniel Povey
d58debdb3f
Use Eve with orthogonalization
2022-05-16 10:34:25 +08:00
Daniel Povey
995371ad95
Move train.py changes to the right dir
2022-05-15 22:17:08 +08:00
Daniel Povey
cee5396058
Orthogonalize every 2k iters
2022-05-15 21:50:40 +08:00
Daniel Povey
bb32556f9e
Add and test reset() function
2022-05-15 16:20:10 +08:00
Daniel Povey
a1dc020270
train scalars slower
2022-05-15 15:44:56 +08:00
Daniel Povey
0989aec741
Replace Eve optimizer with Abel.
2022-05-15 14:33:08 +08:00
Daniel Povey
5e3bf4ce5a
Remove some debug code
2022-05-15 14:22:12 +08:00
Daniel Povey
6306f24430
Some changes to algorithm; more diagnostics printing
2022-05-15 14:12:23 +08:00
Daniel Povey
67c402a369
Add some debugging/diagnostic code
2022-05-15 13:28:00 +08:00
Daniel Povey
747960677e
Prob. 1st working version of Abel
2022-05-15 10:13:06 +08:00
Daniel Povey
a1c6bae5d6
Init pruned_transducer_stateless4b as copy of 4
2022-05-13 22:48:17 +08:00
Daniel Povey
993b36e0b8
Merge remote-tracking branch 'upstream/master' into knowledge_base_1b_L2_ng
2022-05-13 22:47:20 +08:00
Fangjun Kuang
0f180b3ce2
Validate that there are no OOV tokens in BPE-based lexicons. ( #359 )
...
* Validate that there are no OOV tokens in BPE-based lexicons.
* Typo fixes.
2022-05-13 14:00:35 +08:00
Daniel Povey
44f4aa5f66
Try to resolve merge issues etc
2022-05-13 11:32:23 +08:00
Daniel Povey
4f933f5413
Merge changes from knowledge_base_1bfast; fix nheads 4->8
2022-05-13 11:26:18 +08:00
Fangjun Kuang
7b7acdf369
Support --iter in export.py ( #360 )
2022-05-13 10:51:44 +08:00
Fangjun Kuang
aeb8986e35
Ignore padding frames during RNN-T decoding. ( #358 )
...
* Ignore padding frames during RNN-T decoding.
* Fix outdated decoding code.
* Minor fixes.
2022-05-13 07:39:14 +08:00
Daniel Povey
c4c9b8cf80
Change configuration from M to L
2022-05-12 16:10:08 +08:00
Fangjun Kuang
bc284e88e6
Run decode.py in GitHub actions. ( #356 )
2022-05-10 14:51:34 +08:00
Daniel Povey
0bf538a4a3
Add negentropy_penalty, on individual dims.
2022-05-10 13:20:10 +08:00
Zengwei Yao
20f092e709
Support decoding with averaged model when using --iter ( #353 )
...
* support decoding with averaged model when using --iter
* minor fix
* monir fix of copyright date
2022-05-07 13:09:11 +08:00
Zengwei Yao
c059ef3169
Keep model_avg on cpu ( #348 )
...
* keep model_avg on cpu
* explicitly convert model_avg to cpu
* minor fix
* remove device convertion for model_avg
* modify usage of the model device in train.py
* change model.device to next(model.parameters()).device for decoding
* assert params.start_epoch>0
* assert params.start_epoch>0, params.start_epoch
2022-05-07 10:42:34 +08:00
Fangjun Kuang
32f05c00e3
Save batch to disk on exception. ( #350 )
2022-05-06 17:49:40 +08:00
Zengwei Yao
00c48ec1f3
Model average ( #344 )
...
* First upload of model average codes.
* minor fix
* update decode file
* update .flake8
* rename pruned_transducer_stateless3 to pruned_transducer_stateless4
* change epoch number counter starting from 1 instead of 0
* minor fix of pruned_transducer_stateless4/train.py
* refactor the checkpoint.py
* minor fix, update docs, and modify the epoch number to count from 1 in the pruned_transducer_stateless4/decode.py
* update author info
* add docs of the scaling in function average_checkpoints_with_averaged_model
2022-05-05 21:20:04 +08:00
Fangjun Kuang
8635fb4334
Fix decoding for gigaspeech in the libri + giga setup. ( #345 )
2022-05-05 20:58:46 +08:00
Daniel Povey
0f7ff7470f
Switch sampling to new C++/CUDA backend
2022-05-05 15:44:04 +08:00
Fangjun Kuang
e1c3e98980
Save batch to disk on OOM. ( #343 )
...
* Save batch to disk on OOM.
* minor fixes
* Fixes after review.
* Fix style issues.
2022-05-05 15:09:23 +08:00
Fangjun Kuang
6af15914fa
Validate generated manifest files. ( #338 )
2022-05-03 07:02:54 +08:00
Fangjun Kuang
6dc2e04462
Update results. ( #340 )
...
* Update results.
* Typo fixes.
2022-04-29 15:49:45 +08:00
Fangjun Kuang
ac84220de9
Modified conformer with multi datasets ( #312 )
...
* Copy files for editing.
* Use librispeech + gigaspeech with modified conformer.
* Support specifying number of workers for on-the-fly feature extraction.
* Feature extraction code for GigaSpeech.
* Combine XL splits lazily during training.
* Fix warnings in decoding.
* Add decoding code for GigaSpeech.
* Fix decoding the gigaspeech dataset.
We have to use the decoder/joiner networks for the GigaSpeech dataset.
* Disable speed perturbe for XL subset.
* Compute the Nbest oracle WER for RNN-T decoding.
* Minor fixes.
* Minor fixes.
* Add results.
* Update results.
* Update CI.
* Update results.
* Fix style issues.
* Update results.
* Fix style issues.
2022-04-29 15:40:30 +08:00
Fangjun Kuang
caab6cfd92
Support specifying iteration number of checkpoints for decoding. ( #336 )
...
See also #289
2022-04-28 14:09:22 +08:00
Daniel Povey
551786b9bd
Merge branch 'model-averaging-shared-params' of https://github.com/csukuangfj/icefall into knowledge_base_1b_merge
2022-04-26 13:18:09 +08:00
Daniel Povey
eba025a6b4
Mess with thresholds for printing
2022-04-26 10:39:35 +08:00
Daniel Povey
3ba081e6d9
Add more custom_fwd,custom_bwd'
2022-04-25 23:58:34 +08:00
Daniel Povey
2c4478b6d1
Fix for half precision
2022-04-25 23:03:34 +08:00
Daniel Povey
e718c7ac88
Remove unnecessary copy
2022-04-25 20:41:00 +08:00
Daniel Povey
f6619a0b20
Remove unnecessary check
2022-04-25 20:37:06 +08:00
Daniel Povey
7d457a7781
Add some diagnostics
2022-04-25 19:34:19 +08:00
Daniel Povey
edaaec09cd
Update backprop of sampling.py to be slightly more efficient.
2022-04-25 19:32:11 +08:00
pehonnet
9a98e6ced6
fix fp16 option in example usage ( #332 )
2022-04-25 18:51:53 +08:00
Daniel Povey
bbfa484196
Decrease model size, baseline is one Fangjun is running..
2022-04-25 17:07:20 +08:00
Daniel Povey
aea116ea25
Change printing-prob, initial scales
2022-04-25 14:02:43 +08:00
Daniel Povey
bb7cb82b04
Some fixes/refactoring, make parameters shared
2022-04-25 13:55:27 +08:00
Daniel Povey
0d40b4617a
Add knowledge-base lookup to model
2022-04-25 13:40:47 +08:00
