mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-09-11 18:14:19 +00:00
Code Issues Packages Projects Releases Wiki Activity

update RESULT.md about pruned_transducer_stateless4

Browse Source
This commit is contained in:
yaozengwei 2022-05-27 18:07:07 +08:00
parent fbbc24f941
commit 525c1ea277
1 changed files with 80 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

80
egs/librispeech/ASR/RESULTS.md
View File

@ -193,6 +193,86 @@ You can find a pretrained model, training logs, decoding logs, and decoding
results at: results at:
<https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless5-narrower-2022-05-13> <https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless5-narrower-2022-05-13>
### LibriSpeech BPE training results (Pruned Transducer 4)
[pruned_transducer_stateless4](./pruned_transducer_stateless4)
This version saves averaged model during training, and decodes with averaged model.
See <https://github.com/k2-fsa/icefall/issues/337> for details about the idea of model averaging.
#### Training on full librispeech
See <https://github.com/k2-fsa/icefall/pull/344>
Using commit `ec0b0e92297cc03fdb09f48cd235e84d2c04156b`.
The WERs are:
| | test-clean | test-other | comment |
|-------------------------------------|------------|------------|-------------------------------------------------------------------------------|
| greedy search (max sym per frame 1) | 2.77 | 6.78 | --epoch 30 --avg 5 --use_averaged_model False |
| greedy search (max sym per frame 1) | 2.72 | 6.67 | --epoch 30 --avg 5 --use_averaged_model True |
| greedy search (max sym per frame 1) | 2.78 | 6.68 | --epoch 30 --avg 10 --use_averaged_model False |
| greedy search (max sym per frame 1) | 2.74 | 6.67 | --epoch 30 --avg 10 --use_averaged_model True |
| fast beam search | 2.73 | 6.65 | --epoch 30 --avg 5 --use_averaged_model False |
| fast beam search | 2.69 | 6.61 | --epoch 30 --avg 5 --use_averaged_model True |
| fast beam search | 2.68 | 6.61 | --epoch 30 --avg 10 --use_averaged_model False |
| fast beam search | 2.7 | 6.57 | --epoch 30 --avg 10 --use_averaged_model True |
| modified beam search | 2.71 | 6.69 | --epoch 30 --avg 5 --use_averaged_model False |
| modified beam search | 2.65 | 6.6 | --epoch 30 --avg 5 --use_averaged_model True |
| modified beam search | 2.69 | 6.61 | --epoch 30 --avg 10 --use_averaged_model False |
| modified beam search | 2.68 | 6.56 | --epoch 30 --avg 10 --use_averaged_model True |
The training command is:
./pruned_transducer_stateless4/train.py \
--world-size 6 \
--num-epochs 30 \
--start-epoch 1 \
--exp-dir pruned_transducer_stateless4/exp \
--full-libri 1 \
--max-duration 300 \
--save-every-n 8000 \
--keep-last-k 20 \
--average-period 100
#### Training on train-clean-100
See <https://github.com/k2-fsa/icefall/pull/344>
Using commit `ec0b0e92297cc03fdb09f48cd235e84d2c04156b`.
The WERs are:
| | test-clean | test-other | comment |
|-------------------------------------|------------|------------|-------------------------------------------------------------------------------|
| greedy search (max sym per frame 1) | 7.13 | 19.31 | --epoch 30 --avg 5 --use_averaged_model False |
| greedy search (max sym per frame 1) | 7.03 | 18.84 | --epoch 30 --avg 5 --use_averaged_model True |
| greedy search (max sym per frame 1) | 7.0 | 18.95 | --epoch 30 --avg 10 --use_averaged_model False |
| greedy search (max sym per frame 1) | 6.92 | 18.65 | --epoch 30 --avg 10 --use_averaged_model True |
| fast beam search | 6.96 | 18.78 | --epoch 30 --avg 5 --use_averaged_model False |
| fast beam search | 6.88 | 18.43 | --epoch 30 --avg 5 --use_averaged_model True |
| fast beam search | 6.82 | 18.47 | --epoch 30 --avg 10 --use_averaged_model False |
| fast beam search | 6.74 | 18.2 | --epoch 30 --avg 10 --use_averaged_model True |
| modified beam search | 6.86 | 18.76 | --epoch 30 --avg 5 --use_averaged_model False |
| modified beam search | 6.83 | 18.31 | --epoch 30 --avg 5 --use_averaged_model True |
| modified beam search | 6.74 | 18.39 | --epoch 30 --avg 10 --use_averaged_model False |
| modified beam search | 6.74 | 18.12 | --epoch 30 --avg 10 --use_averaged_model True |
The training command is:
./pruned_transducer_stateless4/train.py \
--world-size 3 \
--num-epochs 30 \
--start-epoch 1 \
--exp-dir pruned_transducer_stateless4/exp \
--full-libri 0 \
--max-duration 300 \
--save-every-n 8000 \
--keep-last-k 20 \
--average-period 100
### LibriSpeech BPE training results (Pruned Stateless Transducer 3, 2022-04-29) ### LibriSpeech BPE training results (Pruned Stateless Transducer 3, 2022-04-29)
[pruned_transducer_stateless3](./pruned_transducer_stateless3) [pruned_transducer_stateless3](./pruned_transducer_stateless3)