luomingshuang/icefall_asr_alimeeting_pruned_transducer_stateless2

Note: This recipe is trained with the codes from this PR https://github.com/k2-fsa/icefall/pull/378 And the SpecAugment codes from this PR https://github.com/lhotse-speech/lhotse/pull/604.

Pre-trained Transducer-Stateless2 models for the Alimeeting dataset with icefall.

The model was trained on the far data of Alimeeting with the scripts in icefall based on the latest version k2.

Training procedure

The main repositories are list below, we will update the training and decoding scripts with the update of version.
k2: https://github.com/k2-fsa/k2 icefall: https://github.com/k2-fsa/icefall lhotse: https://github.com/lhotse-speech/lhotse

• Install k2 and lhotse, k2 installation guide refers to https://k2.readthedocs.io/en/latest/installation/index.html, lhotse refers to https://lhotse.readthedocs.io/en/latest/getting-started.html#installation. I think the latest version would be ok. And please also install the requirements listed in icefall.
• Clone icefall(https://github.com/k2-fsa/icefall) and check to the commit showed above.

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

• Preparing data.

cd egs/alimeeting/ASR
bash ./prepare.sh

• Training

export CUDA_VISIBLE_DEVICES="0,1,2,3"
./pruned_transducer_stateless2/train.py \
                  --world-size 4 \
                  --num-epochs 30 \
                  --start-epoch 0 \
                  --exp-dir pruned_transducer_stateless2/exp \
                  --lang-dir data/lang_char \
                  --max-duration 220

Evaluation results

The decoding results (WER%) on Alimeeting(eval and test) are listed below, we got this result by averaging models from epoch 12 to 29. The WERs are

	eval	test	comment
greedy search	31.77	34.66	--epoch 29, --avg 18, --max-duration 100
modified beam search (beam size 4)	30.38	33.02	--epoch 29, --avg 18, --max-duration 100
fast beam search (set as default)	31.39	34.25	--epoch 29, --avg 18, --max-duration 1500