Skip to content

Latest commit

 

History

History
74 lines (53 loc) · 2.41 KB

RESULTS.md

File metadata and controls

74 lines (53 loc) · 2.41 KB
Raw

Results

TIMIT training results (Tdnn_LSTM_CTC)

2021-11-16

(Mingshuang Luo): Result of k2-fsa#114

TensorBoard log is available at https://tensorboard.dev/experiment/qhA1o025Q322kO34SlhWzg/#scalars

Pretrained model is available at https://huggingface.co/luomingshuang/icefall_asr_timit_tdnn_lstm_ctc

The best decoding results (PER) are listed below, we got this results by averaging models from epoch 16 to 25, and using whole-lattice-rescoring with lm_scale equals to 0.08.

TEST
PER 19.71%

You can use the following commands to reproduce our results:

git clone https://github.com/k2-fsa/icefall
cd icefall

cd egs/timit/ASR
./prepare.sh

export CUDA_VISIBLE_DEVICES="0"
python tdnn_lstm_ctc/train.py --bucketing-sampler True \
                              --concatenate-cuts False \
                              --max-duration 200 \
                              --world-size 1 \
                              --lang-dir data/lang_phone

python tdnn_lstm_ctc/decode.py --epoch 25 \
                               --avg 10 \
                               --max-duration 20 \
                               --lang-dir data/lang_phone

TIMIT training results (Tdnn_LiGRU_CTC)

2021-11-16

(Mingshuang Luo): Result of phone based Tdnn_LiGRU_CTC model, k2-fsa#114

TensorBoard log is available at https://tensorboard.dev/experiment/IlQxeq5vQJ2SEVP94Y5fyg/#scalars

Pretrained model is available at https://huggingface.co/luomingshuang/icefall_asr_timit_tdnn_ligru_ctc

The best decoding results (PER) are listed below, we got this results by averaging models from epoch 9 to 25, and using whole-lattice-rescoring decoding method with lm_scale equals to 0.1.

TEST
PER 17.66%

You can use the following commands to reproduce our results:

git clone https://github.com/k2-fsa/icefall
cd icefall

cd egs/timit/ASR
./prepare.sh

export CUDA_VISIBLE_DEVICES="0"
python tdnn_ligru_ctc/train.py --bucketing-sampler True \
                              --concatenate-cuts False \
                              --max-duration 200 \
                              --world-size 1 \
                              --lang-dir data/lang_phone

python tdnn_ligru_ctc/decode.py --epoch 25 \
                               --avg 17 \
                               --max-duration 20 \
                               --lang-dir data/lang_phone