Results

zipformer (zipformer + pruned stateless transducer + CTC)

See #1111 for more details.

Non-streaming

normal-scaled model, number of model parameters: 65805511, i.e., 65.81 M

The tensorboard log can be found at https://tensorboard.dev/experiment/Lo3Qlad7TP68ulM2K0ixgQ/

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/Zengwei/icefall-asr-librispeech-zipformer-transducer-ctc-2023-06-13

You can use https://github.com/k2-fsa/sherpa to deploy it.

Results of the CTC head:

decoding method	test-clean	test-other	comment
ctc-decoding	2.40	5.66	--epoch 40 --avg 16
1best	2.46	5.11	--epoch 40 --avg 16
nbest	2.46	5.11	--epoch 40 --avg 16
nbest-rescoring	2.37	4.93	--epoch 40 --avg 16
whole-lattice-rescoring	2.37	4.88	--epoch 40 --avg 16

The training command is:

export CUDA_VISIBLE_DEVICES="0,1,2,3"
./zipformer/train.py \
  --world-size 4 \
  --num-epochs 40 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir zipformer/exp-ctc-rnnt \
  --causal 0 \
  --use-transducer 1 \
  --use-ctc 1 \
  --ctc-loss-scale 0.2 \
  --full-libri 1 \
  --max-duration 1000

The decoding command is:

export CUDA_VISIBLE_DEVICES="0"
for m in ctc-decoding 1best nbest nbest-rescoring whole-lattice-rescoring; do
  ./zipformer/ctc_decode.py \
      --epoch 40 \
      --avg 16 \
      --exp-dir zipformer/exp-ctc-rnnt \
      --use-transducer 1 \
      --use-ctc 1 \
      --max-duration 300 \
      --causal 0 \
      --num-paths 100 \
      --nbest-scale 1.0 \
      --hlg-scale 0.6 \
      --decoding-method $m
done

zipformer (zipformer + pruned stateless transducer)

See #1058 for more details.

zipformer

Non-streaming

normal-scaled model, number of model parameters: 65549011, i.e., 65.55 M

The tensorboard log can be found at https://tensorboard.dev/experiment/R2DT9Ju4QiadC4e2ioKh5A/

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/Zengwei/icefall-asr-librispeech-zipformer-2023-05-15

You can use https://github.com/k2-fsa/sherpa to deploy it.

decoding method	test-clean	test-other	comment
greedy_search	2.27	5.1	--epoch 30 --avg 9
modified_beam_search	2.25	5.06	--epoch 30 --avg 9
fast_beam_search	2.25	5.04	--epoch 30 --avg 9
greedy_search	2.23	4.96	--epoch 40 --avg 16
modified_beam_search	2.21	4.91	--epoch 40 --avg 16
fast_beam_search	2.24	4.93	--epoch 40 --avg 16
greedy_search	2.22	4.87	--epoch 50 --avg 25
modified_beam_search	2.21	4.79	--epoch 50 --avg 25
fast_beam_search	2.21	4.82	--epoch 50 --avg 25
modified_beam_search_shallow_fusion	2.01	4.37	--epoch 40 --avg 16 --beam-size 12 --lm-scale 0.3
modified_beam_search_LODR	1.94	4.17	--epoch 40 --avg 16 --beam-size 12 --lm-scale 0.52 --LODR-scale -0.26
modified_beam_search_rescore	2.04	4.39	--epoch 40 --avg 16 --beam-size 12
modified_beam_search_rescore_LODR	2.01	4.33	--epoch 40 --avg 16 --beam-size 12

The training command is:

export CUDA_VISIBLE_DEVICES="0,1,2,3"
./zipformer/train.py \
  --world-size 4 \
  --num-epochs 50 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir zipformer/exp \
  --causal 0 \
  --full-libri 1 \
  --max-duration 1000

The decoding command is:

export CUDA_VISIBLE_DEVICES="0"
for m in greedy_search modified_beam_search fast_beam_search; do
  ./zipformer/decode.py \
    --epoch 50 \
    --avg 25 \
    --use-averaged-model 1 \
    --exp-dir ./zipformer/exp \
    --max-duration 600 \
    --decoding-method $m
done

To decode with external language models, please refer to the documentation here.

small-scaled model, number of model parameters: 23285615, i.e., 23.3 M

The tensorboard log can be found at https://tensorboard.dev/experiment/M9C8cYPWSN2MVBYaBIX3EQ/

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/Zengwei/icefall-asr-librispeech-zipformer-small-2023-05-16

You can use https://github.com/k2-fsa/sherpa to deploy it.

decoding method	test-clean	test-other	comment
greedy_search	2.64	6.14	--epoch 30 --avg 8
modified_beam_search	2.6	6.01	--epoch 30 --avg 8
fast_beam_search	2.62	6.06	--epoch 30 --avg 8
greedy_search	2.49	5.91	--epoch 40 --avg 13
modified_beam_search	2.46	5.83	--epoch 40 --avg 13
fast_beam_search	2.46	5.87	--epoch 40 --avg 13
greedy_search	2.46	5.86	--epoch 50 --avg 23
modified_beam_search	2.42	5.73	--epoch 50 --avg 23
fast_beam_search	2.46	5.78	--epoch 50 --avg 23

The training command is:

export CUDA_VISIBLE_DEVICES="0,1"
./zipformer/train.py \
  --world-size 2 \
  --num-epochs 50 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir zipformer/exp-small \
  --causal 0 \
  --num-encoder-layers 2,2,2,2,2,2 \
  --feedforward-dim 512,768,768,768,768,768 \
  --encoder-dim 192,256,256,256,256,256 \
  --encoder-unmasked-dim 192,192,192,192,192,192 \
  --base-lr 0.04 \
  --full-libri 1 \
  --max-duration 1500

The decoding command is:

export CUDA_VISIBLE_DEVICES="0"
for m in greedy_search modified_beam_search fast_beam_search; do
  ./zipformer/decode.py \
    --epoch 50 \
    --avg 23 \
    --exp-dir zipformer/exp-small \
    --max-duration 600 \
    --causal 0 \
    --decoding-method $m \
    --num-encoder-layers 2,2,2,2,2,2 \
    --feedforward-dim 512,768,768,768,768,768 \
    --encoder-dim 192,256,256,256,256,256 \
    --encoder-unmasked-dim 192,192,192,192,192,192
done

large-scaled model, number of model parameters: 148439574, i.e., 148.4 M

The tensorboard log can be found at https://tensorboard.dev/experiment/C5ZPE5u1So2ZwhYLKW0FVg/

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/Zengwei/icefall-asr-librispeech-zipformer-large-2023-05-16

You can use https://github.com/k2-fsa/sherpa to deploy it.

decoding method	test-clean	test-other	comment
greedy_search	2.12	4.91	--epoch 30 --avg 9
modified_beam_search	2.11	4.9	--epoch 30 --avg 9
fast_beam_search	2.13	4.93	--epoch 30 --avg 9
greedy_search	2.12	4.8	--epoch 40 --avg 13
modified_beam_search	2.11	4.7	--epoch 40 --avg 13
fast_beam_search	2.13	4.78	--epoch 40 --avg 13
greedy_search	2.08	4.69	--epoch 50 --avg 30
modified_beam_search	2.06	4.63	--epoch 50 --avg 30
fast_beam_search	2.09	4.68	--epoch 50 --avg 30

The training command is:

export CUDA_VISIBLE_DEVICES="0,1,2,3"
./zipformer/train.py \
  --world-size 4 \
  --num-epochs 50 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir zipformer/exp-large \
  --causal 0 \
  --num-encoder-layers 2,2,4,5,4,2 \
  --feedforward-dim 512,768,1536,2048,1536,768 \
  --encoder-dim 192,256,512,768,512,256 \
  --encoder-unmasked-dim 192,192,256,320,256,192 \
  --full-libri 1 \
  --max-duration 1000

The decoding command is:

export CUDA_VISIBLE_DEVICES="0"
for m in greedy_search modified_beam_search fast_beam_search; do
  ./zipformer/decode.py \
    --epoch 50 \
    --avg 30 \
    --exp-dir zipformer/exp-large \
    --max-duration 600 \
    --causal 0 \
    --decoding-method $m \
    --num-encoder-layers 2,2,4,5,4,2 \
    --feedforward-dim 512,768,1536,2048,1536,768 \
    --encoder-dim 192,256,512,768,512,256 \
    --encoder-unmasked-dim 192,192,256,320,256,192
done

large-scaled model, number of model parameters: 148439574, i.e., 148.4 M, trained on 8 80G-A100 GPUs

The tensorboard log can be found at https://tensorboard.dev/experiment/95TdNyEuQXaWK2PzFpD9yg/

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/Zengwei/icefall-asr-librispeech-zipformer-large-2023-10-26-8-a100

You can use https://github.com/k2-fsa/sherpa to deploy it.

decoding method	test-clean	test-other	comment
greedy_search	2.00	4.47	--epoch 174 --avg 172
modified_beam_search	2.00	4.38	--epoch 174 --avg 172
fast_beam_search	2.00	4.42	--epoch 174 --avg 172

The training command is:

export CUDA_VISIBLE_DEVICES="0,1,2,3"
./zipformer/train.py \
  --world-size 8 \
  --num-epochs 174 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir zipformer/exp-large \
  --causal 0 \
  --num-encoder-layers 2,2,4,5,4,2 \
  --feedforward-dim 512,768,1536,2048,1536,768 \
  --encoder-dim 192,256,512,768,512,256 \
  --encoder-unmasked-dim 192,192,256,320,256,192 \
  --full-libri 1 \
  --max-duration 2200

The decoding command is:

export CUDA_VISIBLE_DEVICES="0"
for m in greedy_search modified_beam_search fast_beam_search; do
  ./zipformer/decode.py \
    --epoch 174 \
    --avg 172 \
    --exp-dir zipformer/exp-large \
    --max-duration 600 \
    --causal 0 \
    --decoding-method $m \
    --num-encoder-layers 2,2,4,5,4,2 \
    --feedforward-dim 512,768,1536,2048,1536,768 \
    --encoder-dim 192,256,512,768,512,256 \
    --encoder-unmasked-dim 192,192,256,320,256,192
done

streaming

normal-scaled model, number of model parameters: 66110931, i.e., 66.11 M

The tensorboard log can be found at https://tensorboard.dev/experiment/9rD0i6rMSWq1O61poWi71A

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/Zengwei/icefall-asr-librispeech-streaming-zipformer-2023-05-17

You can use https://github.com/k2-fsa/sherpa to deploy it.

decoding method	chunk size	test-clean	test-other	decoding mode	comment
greedy_search	320ms	3.06	7.81	simulated streaming	--epoch 30 --avg 8 --chunk-size 16 --left-context-frames 128
greedy_search	320ms	3.06	7.79	chunk-wise	--epoch 30 --avg 8 --chunk-size 16 --left-context-frames 128
modified_beam_search	320ms	3.01	7.69	simulated streaming	--epoch 30 --avg 8 --chunk-size 16 --left-context-frames 128
modified_beam_search	320ms	3.05	7.69	chunk-wise	--epoch 30 --avg 8 --chunk-size 16 --left-context-frames 128
fast_beam_search	320ms	3.04	7.68	simulated streaming	--epoch 30 --avg 8 --chunk-size 16 --left-context-frames 128
fast_beam_search	320ms	3.07	7.69	chunk-wise	--epoch 30 --avg 8 --chunk-size 16 --left-context-frames 128
greedy_search	640ms	2.81	7.15	simulated streaming	--epoch 30 --avg 8 --chunk-size 32 --left-context-frames 256
greedy_search	640ms	2.84	7.16	chunk-wise	--epoch 30 --avg 8 --chunk-size 32 --left-context-frames 256
modified_beam_search	640ms	2.79	7.05	simulated streaming	--epoch 30 --avg 8 --chunk-size 32 --left-context-frames 256
modified_beam_search	640ms	2.81	7.11	chunk-wise	--epoch 30 --avg 8 --chunk-size 32 --left-context-frames 256
fast_beam_search	640ms	2.84	7.04	simulated streaming	--epoch 30 --avg 8 --chunk-size 32 --left-context-frames 256
fast_beam_search	640ms	2.83	7.1	chunk-wise	--epoch 30 --avg 8 --chunk-size 32 --left-context-frames 256

Note: For decoding mode, simulated streaming indicates feeding full utterance during decoding using decode.py, while chunk-size indicates feeding certain number of frames at each time using streaming_decode.py.

The training command is:

export CUDA_VISIBLE_DEVICES="0,1,2,3"
./zipformer/train.py \
  --world-size 4 \
  --num-epochs 40 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir zipformer/exp-causal \
  --causal 1 \
  --full-libri 1 \
  --max-duration 1000

The simulated streaming decoding command is:

export CUDA_VISIBLE_DEVICES="0"
for m in greedy_search modified_beam_search fast_beam_search; do
  ./zipformer/decode.py \
    --epoch 30 \
    --avg 8 \
    --use-averaged-model 1 \
    --exp-dir ./zipformer/exp-causal \
    --causal 1 \
    --chunk-size 16 \
    --left-context-frames 128 \
    --max-duration 600 \
    --decoding-method $m
done

The chunk-wise streaming decoding command is:

export CUDA_VISIBLE_DEVICES="0"
for m in greedy_search modified_beam_search fast_beam_search; do
  ./zipformer/streaming_decode.py \
    --epoch 30 \
    --avg 8 \
    --use-averaged-model 1 \
    --exp-dir ./zipformer/exp-causal \
    --causal 1 \
    --chunk-size 16 \
    --left-context-frames 128 \
    --num-decode-streams 2000 \
    --decoding-method $m
done

Zipformer CTC

zipformer_ctc

See #941 for more details.

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/desh2608/icefall-asr-librispeech-zipformer-ctc

Number of model parameters: 86083707, i.e., 86.08 M

decoding method	test-clean	test-other	comment
ctc-decoding	2.50	5.86	--epoch 30 --avg 9
whole-lattice-rescoring	2.44	5.38	--epoch 30 --avg 9
attention-rescoring	2.35	5.16	--epoch 30 --avg 9
1best	2.01	4.61	--epoch 30 --avg 9

The training commands are:

export CUDA_VISIBLE_DEVICES="0,1,2,3"

./zipformer_ctc/train.py \
  --world-size 4 \
  --num-epochs 30 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir zipformer_ctc/exp \
  --full-libri 1 \
  --max-duration 1000 \
  --master-port 12345

The tensorboard log can be found at: https://tensorboard.dev/experiment/IjPSJjHOQFKPYA5Z0Vf8wg

The decoding command is:

./zipformer_ctc/decode.py \
  --epoch 30 --avg 9 --use-averaged-model True \
  --exp-dir zipformer_ctc/exp \
  --lang-dir data/lang_bpe_500 \
  --lm-dir data/lm \
  --method ctc-decoding

pruned_transducer_stateless7 (Fine-tune with mux)

See #1059 for more details.

pruned_transducer_stateless7

The tensorboard log can be found at https://tensorboard.dev/experiment/MaNDZfO7RzW2Czzf3R2ZRA/

You can find the pretrained model and bpe model needed for fine-tuning at: https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless7-2022-11-11

You can find a fine-tuned model, fine-tuning logs, decoding logs, and decoding results at: https://huggingface.co/yfyeung/icefall-asr-finetune-mux-pruned_transducer_stateless7-2023-05-19

You can use https://github.com/k2-fsa/sherpa to deploy it.

Number of model parameters: 70369391, i.e., 70.37 M

decoding method	dev	test	test-clean	test-other	comment
greedy_search	14.27	14.22	2.08	4.79	--epoch 20 --avg 5
modified_beam_search	14.22	14.08	2.06	4.72	--epoch 20 --avg 5
fast_beam_search	14.23	14.17	2.08	4.09	--epoch 20 --avg 5

The training commands are:

export CUDA_VISIBLE_DEVICES="0,1"

./pruned_transducer_stateless7/finetune.py \
  --world-size 2 \
  --num-epochs 20 \
  --start-epoch 1 \
  --exp-dir pruned_transducer_stateless7/exp_giga_finetune \
  --subset S \
  --use-fp16 1 \
  --base-lr 0.005 \
  --lr-epochs 100 \
  --lr-batches 100000 \
  --bpe-model icefall-asr-librispeech-pruned-transducer-stateless7-2022-11-11/data/lang_bpe_500/bpe.model \
  --do-finetune True \
  --use-mux True \
  --finetune-ckpt icefall-asr-librispeech-pruned-transducer-stateless7-2022-11-11/exp/pretrain.pt \
  --max-duration 500

The decoding commands are:

# greedy_search
./pruned_transducer_stateless7/decode.py \
    --epoch 20 \
    --avg 5 \
    --use-averaged-model 1 \
    --exp-dir ./pruned_transducer_stateless7/exp_giga_finetune \
    --max-duration 600 \
    --decoding-method greedy_search

# modified_beam_search
./pruned_transducer_stateless7/decode.py \
    --epoch 20 \
    --avg 5 \
    --use-averaged-model 1 \
    --exp-dir ./pruned_transducer_stateless7/exp_giga_finetune \
    --max-duration 600 \
    --decoding-method modified_beam_search \
    --beam-size 4

# fast_beam_search
./pruned_transducer_stateless7/decode.py \
    --epoch 20 \
    --avg 5 \
    --use-averaged-model 1 \
    --exp-dir ./pruned_transducer_stateless7/exp_giga_finetune \
    --max-duration 600 \
    --decoding-method fast_beam_search \
    --beam 20.0 \
    --max-contexts 8 \
    --max-states 64

pruned_transducer_stateless7 (zipformer + multidataset(LibriSpeech + GigaSpeech + CommonVoice 13.0))

See #1010 for more details.

pruned_transducer_stateless7

The tensorboard log can be found at https://tensorboard.dev/experiment/SwdJoHgZSZWn8ph9aJLb8g/

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/yfyeung/icefall-asr-multidataset-pruned_transducer_stateless7-2023-05-04

You can use https://github.com/k2-fsa/sherpa to deploy it.

Number of model parameters: 70369391, i.e., 70.37 M

decoding method	test-clean	test-other	comment
greedy_search	1.91	4.06	--epoch 30 --avg 7
modified_beam_search	1.90	3.99	--epoch 30 --avg 7
fast_beam_search	1.90	3.98	--epoch 30 --avg 7

The training commands are:

export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"

./pruned_transducer_stateless7/train.py \
  --world-size 8 \
  --num-epochs 30 \
  --use-multidataset 1 \
  --use-fp16 1 \
  --max-duration 750 \
  --exp-dir pruned_transducer_stateless7/exp

The decoding commands are:

# greedy_search
./pruned_transducer_stateless7/decode.py \
    --epoch 30 \
    --avg 7 \
    --use-averaged-model 1 \
    --exp-dir ./pruned_transducer_stateless7/exp \
    --max-duration 600 \
    --decoding-method greedy_search

# modified_beam_search
./pruned_transducer_stateless7/decode.py \
    --epoch 30 \
    --avg 7 \
    --use-averaged-model 1 \
    --exp-dir ./pruned_transducer_stateless7/exp \
    --max-duration 600 \
    --decoding-method modified_beam_search \
    --beam-size 4

# fast_beam_search
./pruned_transducer_stateless7/decode.py \
    --epoch 30 \
    --avg 7 \
    --use-averaged-model 1 \
    --exp-dir ./pruned_transducer_stateless7/exp \
    --max-duration 600 \
    --decoding-method fast_beam_search \
    --beam 20.0 \
    --max-contexts 8 \
    --max-states 64

Streaming Zipformer-Transducer (Pruned Stateless Transducer + Streaming Zipformer + Multi-Dataset)

pruned_transducer_stateless7_streaming_multi

See #984 for more details.

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/marcoyang/icefall-libri-giga-pruned-transducer-stateless7-streaming-2023-04-04

Number of model parameters: 70369391, i.e., 70.37 M

training on full librispeech + full gigaspeech (with giga_prob=0.9)

The WERs are:

decoding method	chunk size	test-clean	test-other	comment	decoding mode
greedy search	320ms	2.43	6.0	--epoch 20 --avg 4	simulated streaming
greedy search	320ms	2.47	6.13	--epoch 20 --avg 4	chunk-wise
fast beam search	320ms	2.43	5.99	--epoch 20 --avg 4	simulated streaming
fast beam search	320ms	2.8	6.46	--epoch 20 --avg 4	chunk-wise
modified beam search	320ms	2.4	5.96	--epoch 20 --avg 4	simulated streaming
modified beam search	320ms	2.42	6.03	--epoch 20 --avg 4	chunk-size
greedy search	640ms	2.26	5.58	--epoch 20 --avg 4	simulated streaming
greedy search	640ms	2.33	5.76	--epoch 20 --avg 4	chunk-wise
fast beam search	640ms	2.27	5.54	--epoch 20 --avg 4	simulated streaming
fast beam search	640ms	2.37	5.75	--epoch 20 --avg 4	chunk-wise
modified beam search	640ms	2.22	5.5	--epoch 20 --avg 4	simulated streaming
modified beam search	640ms	2.25	5.69	--epoch 20 --avg 4	chunk-size

The model also has good WERs on GigaSpeech. The following WERs are achieved on GigaSpeech test and dev sets:

decoding method	chunk size	dev	test	comment	decoding mode
greedy search	320ms	12.08	11.98	--epoch 20 --avg 4	simulated streaming
greedy search	640ms	11.66	11.71	--epoch 20 --avg 4	simulated streaming
modified beam search	320ms	11.95	11.83	--epoch 20 --avg 4	simulated streaming
modified beam search	320ms	11.65	11.56	--epoch 20 --avg 4	simulated streaming

Note: simulated streaming indicates feeding full utterance during decoding using decode.py, while chunk-size indicates feeding certain number of frames at each time using streaming_decode.py.

The training command is:

./pruned_transducer_stateless7_streaming_multi/train.py \
  --world-size 4 \
  --num-epochs 20 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir pruned_transducer_stateless7_streaming_multi/exp \
  --full-libri 1 \
  --giga-prob 0.9 \
  --max-duration 750 \
  --master-port 12345

The tensorboard log can be found at https://tensorboard.dev/experiment/G4yDMLXGQXexf41i4MA2Tg/#scalars

The simulated streaming decoding command (e.g., chunk-size=320ms) is:

for m in greedy_search fast_beam_search modified_beam_search; do
  ./pruned_transducer_stateless7_streaming_multi/decode.py \
    --epoch 20 \
    --avg 4 \
    --exp-dir ./pruned_transducer_stateless7_streaming_multi/exp \
    --max-duration 600 \
    --decode-chunk-len 32 \
    --right-padding 64 \
    --decoding-method $m
done

The streaming chunk-size decoding command (e.g., chunk-size=320ms) is:

for m in greedy_search modified_beam_search fast_beam_search; do
  ./pruned_transducer_stateless7_streaming_multi/streaming_decode.py \
    --epoch 20 \
    --avg 4 \
    --exp-dir ./pruned_transducer_stateless7_streaming_multi/exp \
    --decoding-method $m \
    --decode-chunk-len 32 \
    --num-decode-streams 2000
done

Smaller model

We also provide a very small version (only 6.1M parameters) of this setup. The training command for the small model is:

./pruned_transducer_stateless7_streaming_multi/train.py \
  --world-size 4 \
  --num-epochs 30 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir pruned_transducer_stateless7_streaming_multi/exp \
  --full-libri 1 \
  --giga-prob 0.9 \
  --num-encoder-layers "2,2,2,2,2" \
  --feedforward-dims "256,256,512,512,256" \
  --nhead "4,4,4,4,4" \
  --encoder-dims "128,128,128,128,128" \
  --attention-dims "96,96,96,96,96" \
  --encoder-unmasked-dims "96,96,96,96,96" \
  --max-duration 1200 \
  --master-port 12345

You can find this pretrained small model and its training logs, decoding logs, and decoding results at: https://huggingface.co/marcoyang/icefall-libri-giga-pruned-transducer-stateless7-streaming-6M-2023-04-03

decoding method	chunk size	test-clean	test-other	comment	decoding mode
greedy search	320ms	5.95	15.03	--epoch 30 --avg 1	simulated streaming
greedy search	640ms	5.61	13.86	--epoch 30 --avg 1	simulated streaming
modified beam search	320ms	5.72	14.34	--epoch 30 --avg 1	simulated streaming
modified beam search	640ms	5.43	13.16	--epoch 30 --avg 1	simulated streaming
fast beam search	320ms	5.88	14.45	--epoch 30 --avg 1	simulated streaming
fast beam search	640ms	5.48	13.31	--epoch 30 --avg 1	simulated streaming

This small model achieves the following WERs on GigaSpeech test and dev sets:

decoding method	chunk size	dev	test	comment	decoding mode
greedy search	320ms	17.57	17.2	--epoch 30 --avg 1	simulated streaming
modified beam search	320ms	16.98	11.98	--epoch 30 --avg 1	simulated streaming

You can find the tensorboard logs at https://tensorboard.dev/experiment/tAc5iXxTQrCQxky5O5OLyw/#scalars.

Streaming Zipformer-Transducer (Pruned Stateless Transducer + Streaming Zipformer)

pruned_transducer_stateless7_streaming

See #787 for more details.

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29

Number of model parameters: 70369391, i.e., 70.37 M

training on full librispeech

The WERs are:

decoding method	chunk size	test-clean	test-other	comment	decoding mode
greedy search	320ms	3.15	8.09	--epoch 30 --avg 9	simulated streaming
greedy search	320ms	3.17	8.24	--epoch 30 --avg 9	chunk-wise
fast beam search	320ms	3.2	8.04	--epoch 30 --avg 9	simulated streaming
fast beam search	320ms	3.36	8.19	--epoch 30 --avg 9	chunk-wise
modified beam search	320ms	3.11	7.93	--epoch 30 --avg 9	simulated streaming
modified beam search	320ms	3.12	8.11	--epoch 30 --avg 9	chunk-size
greedy search	640ms	2.97	7.5	--epoch 30 --avg 9	simulated streaming
greedy search	640ms	2.98	7.67	--epoch 30 --avg 9	chunk-wise
fast beam search	640ms	3.02	7.47	--epoch 30 --avg 9	simulated streaming
fast beam search	640ms	2.96	7.61	--epoch 30 --avg 9	chunk-wise
modified beam search	640ms	2.94	7.36	--epoch 30 --avg 9	simulated streaming
modified beam search	640ms	2.95	7.53	--epoch 30 --avg 9	chunk-size

Note: simulated streaming indicates feeding full utterance during decoding using decode.py, while chunk-size indicates feeding certain number of frames at each time using streaming_decode.py.

The training command is:

./pruned_transducer_stateless7_streaming/train.py \
  --world-size 4 \
  --num-epochs 30 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir pruned_transducer_stateless7_streaming/exp \
  --full-libri 1 \
  --max-duration 750 \
  --master-port 12345

The tensorboard log can be found at https://tensorboard.dev/experiment/A46UpqEWQWS7oDi5VcQ8rg/

The simulated streaming decoding command (e.g., chunk-size=320ms) is:

for m in greedy_search fast_beam_search modified_beam_search; do
  ./pruned_transducer_stateless7_streaming/decode.py \
    --epoch 30 \
    --avg 9 \
    --exp-dir ./pruned_transducer_stateless7_streaming/exp \
    --max-duration 600 \
    --decode-chunk-len 32 \
    --decoding-method $m
done

The streaming chunk-size decoding command (e.g., chunk-size=320ms) is:

for m in greedy_search modified_beam_search fast_beam_search; do
  ./pruned_transducer_stateless7_streaming/streaming_decode.py \
    --epoch 30 \
    --avg 9 \
    --exp-dir ./pruned_transducer_stateless7_streaming/exp \
    --decoding-method $m \
    --decode-chunk-len 32 \
    --num-decode-streams 2000
done

We also support decoding with neural network LMs. After combining with language models, the WERs are

decoding method	chunk size	test-clean	test-other	comment	decoding mode
`modified_beam_search`	320ms	3.11	7.93	--epoch 30 --avg 9	simulated streaming
`modified_beam_search_lm_shallow_fusion`	320ms	2.58	6.65	--epoch 30 --avg 9	simulated streaming
`modified_beam_search_lm_rescore`	320ms	2.59	6.86	--epoch 30 --avg 9	simulated streaming
`modified_beam_search_lm_rescore_LODR`	320ms	2.52	6.73	--epoch 30 --avg 9	simulated streaming

Please use the following command for modified_beam_search_lm_shallow_fusion:

for lm_scale in $(seq 0.15 0.01 0.38); do
    for beam_size in 4 8 12; do
        ./pruned_transducer_stateless7_streaming/decode.py \
            --epoch 99 \
            --avg 1 \
            --use-averaged-model False \
            --beam-size $beam_size \
            --exp-dir ./pruned_transducer_stateless7_streaming/exp-large-LM \
            --max-duration 600 \
            --decode-chunk-len 32 \
            --decoding-method modified_beam_search_lm_shallow_fusion \
            --use-shallow-fusion 1 \
            --lm-type rnn \
            --lm-exp-dir rnn_lm/exp \
            --lm-epoch 99 \
            --lm-scale $lm_scale \
            --lm-avg 1 \
            --rnn-lm-embedding-dim 2048 \
            --rnn-lm-hidden-dim 2048 \
            --rnn-lm-num-layers 3 \
            --lm-vocab-size 500
    done
done

Please use the following command for modified_beam_search_lm_rescore:

./pruned_transducer_stateless7_streaming/decode.py \
    --epoch 30 \
    --avg 9 \
    --use-averaged-model True \
    --beam-size 8 \
    --exp-dir ./pruned_transducer_stateless7_streaming/exp \
    --max-duration 600 \
    --decode-chunk-len 32 \
    --decoding-method modified_beam_search_lm_rescore \
    --use-shallow-fusion 0 \
    --lm-type rnn \
    --lm-exp-dir rnn_lm/exp \
    --lm-epoch 99 \
    --lm-avg 1 \
    --rnn-lm-embedding-dim 2048 \
    --rnn-lm-hidden-dim 2048 \
    --rnn-lm-num-layers 3 \
    --lm-vocab-size 500

Please use the following command for modified_beam_search_lm_rescore_LODR:

./pruned_transducer_stateless7_streaming/decode.py \
    --epoch 30 \
    --avg 9 \
    --use-averaged-model True \
    --beam-size 8 \
    --exp-dir ./pruned_transducer_stateless7_streaming/exp \
    --max-duration 600 \
    --decode-chunk-len 32 \
    --decoding-method modified_beam_search_lm_rescore_LODR \
    --use-shallow-fusion 0 \
    --lm-type rnn \
    --lm-exp-dir rnn_lm/exp \
    --lm-epoch 99 \
    --lm-avg 1 \
    --rnn-lm-embedding-dim 2048 \
    --rnn-lm-hidden-dim 2048 \
    --rnn-lm-num-layers 3 \
    --lm-vocab-size 500 \
    --tokens-ngram 2 \
    --backoff-id 500

A well-trained RNNLM can be found here: https://huggingface.co/ezerhouni/icefall-librispeech-rnn-lm/tree/main. The bi-gram used in LODR decoding can be found here: https://huggingface.co/marcoyang/librispeech_bigram.

Smaller model

A smaller model (~20M params) is also available with configuration based on this comment. The WERs are:

decoding method	chunk size	test-clean	test-other	comment	decoding mode
greedy search	320ms	3.94	9.79	--epoch 30 --avg 9	simulated streaming
modified beam search	320ms	3.88	9.53	--epoch 30 --avg 9	simulated streaming

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/desh2608/icefall-asr-librispeech-pruned-transducer-stateless7-streaming-small

zipformer_mmi (zipformer with mmi loss)

See #746 for more details.

zipformer_mmi

The tensorboard log can be found at https://tensorboard.dev/experiment/xyOZUKpEQm62HBIlUD4uPA/

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/Zengwei/icefall-asr-librispeech-zipformer-mmi-2022-12-08

Number of model parameters: 69136519, i.e., 69.14 M

	test-clean	test-other	comment
1best	2.54	5.65	--epoch 30 --avg 10
nbest	2.54	5.66	--epoch 30 --avg 10
nbest-rescoring-LG	2.49	5.42	--epoch 30 --avg 10
nbest-rescoring-3-gram	2.52	5.62	--epoch 30 --avg 10
nbest-rescoring-4-gram	2.5	5.51	--epoch 30 --avg 10

The training commands are:

export CUDA_VISIBLE_DEVICES="0,1,2,3"

./zipformer_mmi/train.py \
  --world-size 4 \
  --master-port 12345 \
  --num-epochs 30 \
  --start-epoch 1 \
  --lang-dir data/lang_bpe_500 \
  --max-duration 500 \
  --full-libri 1 \
  --use-fp16 1 \
  --exp-dir zipformer_mmi/exp

The decoding commands for the transducer branch are:

export CUDA_VISIBLE_DEVICES="5"

for m in nbest nbest-rescoring-LG nbest-rescoring-3-gram nbest-rescoring-4-gram; do
  ./zipformer_mmi/decode.py \
    --epoch 30 \
    --avg 10 \
    --exp-dir ./zipformer_mmi/exp/ \
    --max-duration 100 \
    --lang-dir data/lang_bpe_500 \
    --nbest-scale 1.2 \
    --hp-scale 1.0 \
    --decoding-method $m
done

pruned_transducer_stateless7_ctc_bs (zipformer with transducer loss and ctc loss using blank skip)

See #730 for more details.

pruned_transducer_stateless7_ctc_bs

The tensorboard log can be found at https://tensorboard.dev/experiment/rrNZ7l83Qu6RKoD7y49wiA/

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/yfyeung/icefall-asr-librispeech-pruned_transducer_stateless7_ctc_bs-2023-01-29

Number of model parameters: 76804822, i.e., 76.80 M

Test on 8-card V100 cluster, with 4-card busy and 4-card idle.

greedy_search

model	test-clean	test-other	decoding time(s)	comment
pruned_transducer_stateless7_ctc_bs	2.28	5.53	48.939	--epoch 30 --avg 13
pruned_transducer_stateless7_ctc	2.24	5.18	91.900	--epoch 30 --avg 8

pruned_transducer_stateless7_ctc_bs applies blank skip both on training and decoding, and pruned_transducer_stateless7_ctc doesn`t apply blank skip.
Applying blank skip both on training and decoding is 1.88 times faster than the model that doesn't apply blank skip without obvious performance loss.

modified_beam_search

model	test-clean	test-other	decoding time(s)	comment
pruned_transducer_stateless7_ctc_bs	2.26	5.44	80.446	--epoch 30 --avg 13
pruned_transducer_stateless7_ctc	2.20	5.12	283.676	--epoch 30 --avg 8

pruned_transducer_stateless7_ctc_bs applies blank skip both on training and decoding, and pruned_transducer_stateless7_ctc doesn`t apply blank skip.
Applying blank skip both on training and decoding is 3.53 times faster than the model that doesn't apply blank skip without obvious performance loss.

The training commands for the model using blank skip (pruned_transducer_stateless7_ctc_bs) are:

export CUDA_VISIBLE_DEVICES="0,1,2,3"

./pruned_transducer_stateless7_ctc_bs/train.py \
  --world-size 4 \
  --num-epochs 30 \
  --full-libri 1 \
  --use-fp16 1 \
  --max-duration 750 \
  --exp-dir pruned_transducer_stateless7_ctc_bs/exp \
  --feedforward-dims  "1024,1024,2048,2048,1024" \
  --ctc-loss-scale 0.2 \
  --master-port 12535

The decoding commands for the transducer branch of the model using blank skip (pruned_transducer_stateless7_ctc_bs) are:

for m in greedy_search modified_beam_search fast_beam_search; do
  for epoch in 30; do
    for avg in 15; do
      ./pruned_transducer_stateless7_ctc_bs/ctc_guide_decode_bs.py \
          --epoch $epoch \
          --avg $avg \
          --use-averaged-model 1 \
          --exp-dir ./pruned_transducer_stateless7_ctc_bs/exp \
          --feedforward-dims  "1024,1024,2048,2048,1024" \
          --max-duration 600 \
          --decoding-method $m
    done
  done
done

The decoding commands for the transducer branch of the model without blank skip (pruned_transducer_stateless7_ctc) are:

for m in greedy_search modified_beam_search fast_beam_search; do
  for epoch in 30; do
    for avg in 15; do
      ./pruned_transducer_stateless7_ctc/decode.py \
          --epoch $epoch \
          --avg $avg \
          --use-averaged-model 1 \
          --exp-dir ./pruned_transducer_stateless7_ctc/exp \
          --feedforward-dims  "1024,1024,2048,2048,1024" \
          --max-duration 600 \
          --decoding-method $m
    done
  done
done

pruned_transducer_stateless7_ctc (zipformer with transducer loss and ctc loss)

See #683 for more details.

pruned_transducer_stateless7_ctc

The tensorboard log can be found at https://tensorboard.dev/experiment/hxlGAhOPToGmRLZFnAzPWw/

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless7-ctc-2022-12-01

Number of model parameters: 70561891, i.e., 70.56 M

	test-clean	test-other	comment
greedy search	2.23	5.19	--epoch 30 --avg 8
modified beam search	2.21	5.12	--epoch 30 --avg 8
fast beam search	2.23	5.18	--epoch 30 --avg 8
ctc decoding	2.48	5.82	--epoch 30 --avg 9
1best	2.43	5.22	--epoch 30 --avg 9
nbest	2.43	5.22	--epoch 30 --avg 9
nbest rescoring	2.34	5.05	--epoch 30 --avg 9
whole lattice rescoring	2.34	5.04	--epoch 30 --avg 9

The training commands are:

export CUDA_VISIBLE_DEVICES="0,1,2,3"

./pruned_transducer_stateless7_ctc/train.py \
  --world-size 4 \
  --num-epochs 30 \
  --full-libri 1 \
  --use-fp16 1 \
  --max-duration 750 \
  --exp-dir pruned_transducer_stateless7_ctc/exp \
  --feedforward-dims  "1024,1024,2048,2048,1024" \
  --ctc-loss-scale 0.2 \
  --master-port 12535

The decoding commands for the transducer branch are:

for m in greedy_search fast_beam_search modified_beam_search ; do
  for epoch in 30; do
    for avg in 8; do
      ./pruned_transducer_stateless7_ctc/decode.py \
          --epoch $epoch \
          --avg $avg \
          --use-averaged-model 1 \
          --exp-dir ./pruned_transducer_stateless7_ctc/exp \
          --feedforward-dims  "1024,1024,2048,2048,1024" \
          --max-duration 600 \
          --decoding-method $m
    done
  done
done

The decoding commands for the ctc branch are:

for m in ctc-decoding nbest nbest-rescoring whole-lattice-rescoring; do
  for epoch in 30; do
    for avg in 9; do
      ./pruned_transducer_stateless7_ctc/ctc_decode.py \
          --epoch $epoch \
          --avg $avg \
          --exp-dir ./pruned_transducer_stateless7_ctc/exp \
          --max-duration 100 \
          --decoding-method $m \
          --hlg-scale 0.6 \
          --lm-dir data/lm
    done
  done
done

LibriSpeech BPE training results (Conformer CTC, supporting delay penalty)

conformer_ctc3

It implements Conformer model training with CTC loss. For streaming mode, it supports symbol delay penalty.

See #669 for more details.

training on full librispeech

This model contains 12 encoder layers. The number of model parameters is 77352694.

The WERs are:

	test-clean	test-other	comment
ctc-decoding	3.09	7.62	--epoch 25 --avg 7
1best	2.87	6.44	--epoch 25 --avg 7
nbest	2.88	6.5	--epoch 25 --avg 7
nbest-rescoring	2.71	6.1	--epoch 25 --avg 7
whole-lattice-rescoring	2.71	6.04	--epoch 25 --avg 7

The training command is:

./conformer_ctc3/train.py \
  --world-size 4 \
  --num-epochs 25 \
  --start-epoch 1 \
  --exp-dir conformer_ctc3/full \
  --full-libri 1 \
  --max-duration 300 \
  --master-port 12345

The tensorboard log can be found at https://tensorboard.dev/experiment/4jbxIQ2SQIaQeRqsR6bOSA

The decoding command using different methods is:

for method in ctc-decoding 1best nbest nbest-rescoring whole-lattice-rescoring; do
  ./conformer_ctc3/decode.py \
    --epoch 25 \
    --avg 7 \
    --exp-dir conformer_ctc3/exp \
    --max-duration 300 \
    --decoding-method $method \
    --manifest-dir data/fbank \
    --lm-dir data/lm \
done

Pretrained models, training logs, decoding logs, and decoding results are available at https://huggingface.co/Zengwei/icefall-asr-librispeech-conformer-ctc3-2022-11-27

The command to train a streaming model with symbol delay penalty is:

./conformer_ctc3/train.py \
  --world-size 4 \
  --num-epochs 30 \
  --start-epoch 1 \
  --exp-dir conformer_ctc3/exp \
  --full-libri 1 \
  --dynamic-chunk-training 1 \
  --causal-convolution 1 \
  --short-chunk-size 25 \
  --num-left-chunks 4 \
  --max-duration 300 \
  --delay-penalty 0.1

To evaluate symbol delay, you should: (1) Generate cuts with word-time alignments:

./local/add_alignment_librispeech.py \
  --alignments-dir data/alignment \
  --cuts-in-dir data/fbank \
  --cuts-out-dir data/fbank_ali

(2) Set the argument "--manifest-dir data/fbank_ali" while decoding. For example:

./conformer_ctc3/decode.py \
  --epoch 25 \
  --avg 7 \
  --exp-dir ./conformer_ctc3/exp \
  --max-duration 300 \
  --decoding-method ctc-decoding \
  --simulate-streaming 1 \
  --causal-convolution 1 \
  --decode-chunk-size 16 \
  --left-context 64 \
  --manifest-dir data/fbank_ali

Note: It supports to calculate symbol delay with following decoding methods:

ctc-greedy-search
ctc-decoding
1best

pruned_transducer_stateless8 (zipformer + multidataset)

See #675 for more details.

pruned_transducer_stateless8

The tensorboard log can be found at https://tensorboard.dev/experiment/3e9AfOcgRwOXpLQlZvHZrQ

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/WeijiZhuang/icefall-asr-librispeech-pruned-transducer-stateless8-2022-12-02

You can use https://github.com/k2-fsa/sherpa to deploy it.

Number of model parameters: 70369391, i.e., 70.37 M

decoding method	test-clean	test-other	comment
greedy_search	1.81	4.18	--epoch 20 --avg 4
fast_beam_search	1.82	4.15	--epoch 20 --avg 4
modified_beam_search	1.78	4.08	--epoch 20 --avg 4
greedy_search	1.84	4.3	--epoch 19 --avg 8
fast_beam_search	1.77	4.25	--epoch 19 --avg 8
modified_beam_search	1.81	4.16	--epoch 19 --avg 8

The training commands are:

export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"

./pruned_transducer_stateless8/train.py \
  --world-size 8 \
  --num-epochs 20 \
  --full-libri 1 \
  --use-fp16 1 \
  --max-duration 750 \
  --exp-dir pruned_transducer_stateless8/exp \
  --feedforward-dims  "1024,1024,2048,2048,1024" \
  --master-port 12535 \
  --giga-prob 0.9

The decoding commands are:

for m in greedy_search fast_beam_search modified_beam_search; do
  for epoch in $(seq 20 -1 10); do
    for avg in $(seq 9 -1 1); do
      ./pruned_transducer_stateless8/decode.py \
          --epoch $epoch \
          --avg $avg \
          --use-averaged-model 1 \
          --exp-dir ./pruned_transducer_stateless8/exp \
          --feedforward-dims "1024,1024,2048,2048,1024" \
          --max-duration 600 \
          --decoding-method $m
    done
  done
done

pruned_transducer_stateless7 (zipformer)

See #672 for more details.

pruned_transducer_stateless7

The tensorboard log can be found at https://tensorboard.dev/experiment/P7vXWqK7QVu1mU9Ene1gGg/

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless7-2022-11-11

You can use https://github.com/k2-fsa/sherpa to deploy it.

Number of model parameters: 70369391, i.e., 70.37 M

	test-clean	test-other	comment
greedy search	2.17	5.23	--epoch 30 --avg 9 --max-duration 600
modified beam search	2.15	5.20	--epoch 30 --avg 9 --max-duration 600
modified beam search + RNNLM shallow fusion	1.99	4.73	--epoch 30 --avg 9 --max-duration 600
modified beam search + TransformerLM shallow fusion	1.94	4.73	--epoch 30 --avg 9 --max-duration 600
modified beam search + RNNLM + LODR	1.91	4.57	--epoch 30 --avg 9 --max-duration 600
modified beam search + TransformerLM + LODR	1.91	4.51	--epoch 30 --avg 9 --max-duration 600
fast beam search	2.15	5.22	--epoch 30 --avg 9 --max-duration 600

The training commands are:

export CUDA_VISIBLE_DEVICES="0,3,6,7"

./pruned_transducer_stateless7/train.py \
  --world-size 4 \
  --num-epochs 30 \
  --full-libri 1 \
  --use-fp16 1 \
  --max-duration 750 \
  --exp-dir pruned_transducer_stateless7/exp \
  --feedforward-dims  "1024,1024,2048,2048,1024" \
  --master-port 12535

The decoding commands are:

for m in greedy_search fast_beam_search modified_beam_search ; do
  for epoch in 30; do
    for avg in 9; do
      ./pruned_transducer_stateless7/decode.py \
          --epoch $epoch \
          --avg $avg \
          --use-averaged-model 1 \
          --exp-dir ./pruned_transducer_stateless7/exp \
          --feedforward-dims  "1024,1024,2048,2048,1024" \
          --max-duration 600 \
          --decoding-method $m
    done
  done
done

Note that a small change is made to the pruned_transducer_stateless7/decoder.py in this PR to address the problem of emitting the first symbol at the very beginning. If you need a model without this issue, please download the model from here: https://huggingface.co/marcoyang/icefall-asr-librispeech-pruned-transducer-stateless7-2023-03-10

LibriSpeech BPE training results (Pruned Stateless LSTM RNN-T + gradient filter)

lstm_transducer_stateless3

It implements LSTM model with mechanisms in reworked model for streaming ASR. Gradient filter is applied inside each lstm module to stabilize the training.

See #564 for more details.

training on full librispeech

This model contains 12 encoder layers (LSTM module + Feedforward module). The number of model parameters is 84689496.

The WERs are:

	test-clean	test-other	comment	decoding mode
greedy search (max sym per frame 1)	3.66	9.51	--epoch 40 --avg 15	simulated streaming
greedy search (max sym per frame 1)	3.66	9.48	--epoch 40 --avg 15	streaming
fast beam search	3.55	9.33	--epoch 40 --avg 15	simulated streaming
fast beam search	3.57	9.25	--epoch 40 --avg 15	streaming
modified beam search	3.55	9.28	--epoch 40 --avg 15	simulated streaming
modified beam search	3.54	9.25	--epoch 40 --avg 15	streaming

Note: simulated streaming indicates feeding full utterance during decoding, while streaming indicates feeding certain number of frames at each time.

The training command is:

./lstm_transducer_stateless3/train.py \
  --world-size 4 \
  --num-epochs 40 \
  --start-epoch 1 \
  --exp-dir lstm_transducer_stateless3/exp \
  --full-libri 1 \
  --max-duration 500 \
  --master-port 12325 \
  --num-encoder-layers 12 \
  --grad-norm-threshold 25.0 \
  --rnn-hidden-size 1024

The tensorboard log can be found at https://tensorboard.dev/experiment/caNPyr5lT8qAl9qKsXEeEQ/

The simulated streaming decoding command using greedy search, fast beam search, and modified beam search is:

for decoding_method in greedy_search fast_beam_search modified_beam_search; do
  ./lstm_transducer_stateless3/decode.py \
    --epoch 40 \
    --avg 15 \
    --exp-dir lstm_transducer_stateless3/exp \
    --max-duration 600 \
    --num-encoder-layers 12 \
    --rnn-hidden-size 1024 \
    --decoding-method $decoding_method \
    --use-averaged-model True \
    --beam 4 \
    --max-contexts 4 \
    --max-states 8 \
    --beam-size 4
done

The streaming decoding command using greedy search, fast beam search, and modified beam search is:

for decoding_method in greedy_search fast_beam_search modified_beam_search; do
  ./lstm_transducer_stateless3/streaming_decode.py \
    --epoch 40 \
    --avg 15 \
    --exp-dir lstm_transducer_stateless3/exp \
    --max-duration 600 \
    --num-encoder-layers 12 \
    --rnn-hidden-size 1024 \
    --decoding-method $decoding_method \
    --use-averaged-model True \
    --beam 4 \
    --max-contexts 4 \
    --max-states 8 \
    --beam-size 4
done

Pretrained models, training logs, decoding logs, and decoding results are available at https://huggingface.co/Zengwei/icefall-asr-librispeech-lstm-transducer-stateless3-2022-09-28

LibriSpeech BPE training results (Pruned Stateless LSTM RNN-T + multi-dataset)

lstm_transducer_stateless2

See #558 for more details.

The WERs are:

	test-clean	test-other	comment
greedy search (max sym per frame 1)	2.78	7.36	--iter 468000 --avg 16
modified_beam_search	2.73	7.15	--iter 468000 --avg 16
modified_beam_search + RNNLM shallow fusion	2.42	6.46	--iter 468000 --avg 16
modified_beam_search + TransformerLM shallow fusion	2.37	6.48	--iter 468000 --avg 16
modified_beam_search + RNNLM + LODR	2.24	5.89	--iter 468000 --avg 16
modified_beam_search + TransformerLM + LODR	2.19	5.90	--iter 468000 --avg 16
fast_beam_search	2.76	7.31	--iter 468000 --avg 16
greedy search (max sym per frame 1)	2.77	7.35	--iter 472000 --avg 18
modified_beam_search	2.75	7.08	--iter 472000 --avg 18
fast_beam_search	2.77	7.29	--iter 472000 --avg 18

The training command is:

#!/usr/bin/env bash

export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"

./lstm_transducer_stateless2/train.py \
  --world-size 8 \
  --num-epochs 35 \
  --start-epoch 1 \
  --full-libri 1 \
  --exp-dir lstm_transducer_stateless2/exp \
  --max-duration 500 \
  --use-fp16 0 \
  --lr-epochs 10 \
  --num-workers 2 \
  --giga-prob 0.9

Note: It was killed manually after getting epoch-18.pt. Also, we resumed training after getting epoch-9.pt.

The tensorboard log can be found at https://tensorboard.dev/experiment/1ziQ2LFmQY2mt4dlUr5dyA/

The decoding command is

for m in greedy_search fast_beam_search modified_beam_search; do
  for iter in 472000; do
    for avg in 8 10 12 14 16 18; do
      ./lstm_transducer_stateless2/decode.py \
        --iter $iter \
        --avg $avg \
        --exp-dir lstm_transducer_stateless2/exp \
        --max-duration 600 \
        --num-encoder-layers 12 \
        --rnn-hidden-size 1024 \
        --decoding-method $m \
        --use-averaged-model True \
        --beam 4 \
        --max-contexts 4 \
        --max-states 8 \
        --beam-size 4
    done
  done
done

You may also decode using shallow fusion with external neural network LM. To do so you need to download a well-trained NN LM: RNN LM: https://huggingface.co/ezerhouni/icefall-librispeech-rnn-lm/tree/main Transformer LM: https://huggingface.co/marcoyang/icefall-librispeech-transformer-lm/tree/main

for iter in 472000; do
    for avg in 8 10 12 14 16 18; do
        ./lstm_transducer_stateless2/decode.py \
                --iter $iter \
                --avg $avg \
                --exp-dir ./lstm_transducer_stateless2/exp \
                --max-duration 600 \
                --decoding-method modified_beam_search_lm_shallow_fusion \
                --use-shallow-fusion 1 \
                --lm-type rnn \
                --lm-exp-dir /ceph-data4/yangxiaoyu/pretrained_models/LM/icefall-librispeech-rnn-lm/exp \
                --lm-epoch 99 \
                --lm-scale $lm_scale \
                --lm-avg 1 \
    done
done

You may also decode using LODR + LM shallow fusion. This decoding method is proposed in https://arxiv.org/pdf/2203.16776.pdf. It subtracts the internal language model score during shallow fusion, which is approximated by a bi-gram model. The bi-gram can be generated by prepare_lm.sh at stage 4, or you may download it from https://huggingface.co/marcoyang/librispeech_bigram.

The decoding command is as follows:

for iter in 472000; do
    for avg in 8 10 12 14 16 18; do
        ./lstm_transducer_stateless2/decode.py \
                --iter $iter \
                --avg $avg \
                --exp-dir ./lstm_transducer_stateless2/exp \
                --max-duration 600 \
                --decoding-method modified_beam_search_LODR \
                --beam 4 \
                --max-contexts 4 \
                --use-shallow-fusion 1 \
                --lm-type rnn \
                --lm-exp-dir /ceph-data4/yangxiaoyu/pretrained_models/LM/icefall-librispeech-rnn-lm/exp \
                --lm-epoch 99 \
                --lm-scale 0.4 \
                --lm-avg 1 \
                --tokens-ngram 2 \
                --ngram-lm-scale -0.16
    done
done

Note that you can also set --lm-type transformer to use transformer LM during LODR. But it will be slower because it has not been optimized. The pre-trained transformer LM is available at https://huggingface.co/marcoyang/icefall-librispeech-transformer-lm/tree/main

Pretrained models, training logs, decoding logs, and decoding results are available at https://huggingface.co/csukuangfj/icefall-asr-librispeech-lstm-transducer-stateless2-2022-09-03

LibriSpeech BPE training results (Pruned Stateless LSTM RNN-T)

lstm_transducer_stateless

It implements LSTM model with mechanisms in reworked model for streaming ASR.

See #479 for more details.

training on full librispeech

This model contains 12 encoder layers (LSTM module + Feedforward module). The number of model parameters is 84689496.

The WERs are:

	test-clean	test-other	comment	decoding mode
greedy search (max sym per frame 1)	3.81	9.73	--epoch 35 --avg 15	simulated streaming
greedy search (max sym per frame 1)	3.78	9.79	--epoch 35 --avg 15	streaming
fast beam search	3.74	9.59	--epoch 35 --avg 15	simulated streaming
fast beam search	3.73	9.61	--epoch 35 --avg 15	streaming
modified beam search	3.64	9.55	--epoch 35 --avg 15	simulated streaming
modified beam search	3.65	9.51	--epoch 35 --avg 15	streaming

Note: simulated streaming indicates feeding full utterance during decoding, while streaming indicates feeding certain number of frames at each time.

The training command is:

./lstm_transducer_stateless/train.py \
  --world-size 4 \
  --num-epochs 35 \
  --start-epoch 1 \
  --exp-dir lstm_transducer_stateless/exp \
  --full-libri 1 \
  --max-duration 500 \
  --master-port 12321 \
  --num-encoder-layers 12 \
  --rnn-hidden-size 1024

The tensorboard log can be found at https://tensorboard.dev/experiment/FWrM20mjTeWo6dTpFYOsYQ/

The simulated streaming decoding command using greedy search, fast beam search, and modified beam search is:

for decoding_method in greedy_search fast_beam_search modified_beam_search; do
  ./lstm_transducer_stateless/decode.py \
    --epoch 35 \
    --avg 15 \
    --exp-dir lstm_transducer_stateless/exp \
    --max-duration 600 \
    --num-encoder-layers 12 \
    --rnn-hidden-size 1024 \
    --decoding-method $decoding_method \
    --use-averaged-model True \
    --beam 4 \
    --max-contexts 4 \
    --max-states 8 \
    --beam-size 4
done

The streaming decoding command using greedy search, fast beam search, and modified beam search is:

for decoding_method in greedy_search fast_beam_search modified_beam_search; do
  ./lstm_transducer_stateless/streaming_decode.py \
    --epoch 35 \
    --avg 15 \
    --exp-dir lstm_transducer_stateless/exp \
    --max-duration 600 \
    --num-encoder-layers 12 \
    --rnn-hidden-size 1024 \
    --decoding-method $decoding_method \
    --use-averaged-model True \
    --beam 4 \
    --max-contexts 4 \
    --max-states 8 \
    --beam-size 4
done

Pretrained models, training logs, decoding logs, and decoding results are available at https://huggingface.co/Zengwei/icefall-asr-librispeech-lstm-transducer-stateless-2022-08-18

LibriSpeech BPE training results (Pruned Stateless Conv-Emformer RNN-T 2)

conv_emformer_transducer_stateless2

It implements Emformer augmented with convolution module and simplified memory bank for streaming ASR. It is modified from torchaudio.

See #440 for more details.

With lower latency setup, training on full librispeech

In this model, the lengths of chunk and right context are 32 frames (i.e., 0.32s) and 8 frames (i.e., 0.08s), respectively.

The WERs are:

	test-clean	test-other	comment	decoding mode
greedy search (max sym per frame 1)	3.5	9.09	--epoch 30 --avg 10	simulated streaming
greedy search (max sym per frame 1)	3.57	9.1	--epoch 30 --avg 10	streaming
fast beam search	3.5	8.91	--epoch 30 --avg 10	simulated streaming
fast beam search	3.54	8.91	--epoch 30 --avg 10	streaming
modified beam search	3.43	8.86	--epoch 30 --avg 10	simulated streaming
modified beam search	3.48	8.88	--epoch 30 --avg 10	streaming

The training command is:

./conv_emformer_transducer_stateless2/train.py \
  --world-size 6 \
  --num-epochs 30 \
  --start-epoch 1 \
  --exp-dir conv_emformer_transducer_stateless2/exp \
  --full-libri 1 \
  --max-duration 280 \
  --master-port 12321 \
  --num-encoder-layers 12 \
  --chunk-length 32 \
  --cnn-module-kernel 31 \
  --left-context-length 32 \
  --right-context-length 8 \
  --memory-size 32

The tensorboard log can be found at https://tensorboard.dev/experiment/W5MpxekiQLSPyM4fe5hbKg/

The simulated streaming decoding command using greedy search is:

./conv_emformer_transducer_stateless2/decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless2/exp \
  --max-duration 300 \
  --num-encoder-layers 12 \
  --chunk-length 32 \
  --cnn-module-kernel 31 \
  --left-context-length 32 \
  --right-context-length 8 \
  --memory-size 32 \
  --decoding-method greedy_search \
  --use-averaged-model True

The simulated streaming decoding command using fast beam search is:

./conv_emformer_transducer_stateless2/decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless2/exp \
  --max-duration 300 \
  --num-encoder-layers 12 \
  --chunk-length 32 \
  --cnn-module-kernel 31 \
  --left-context-length 32 \
  --right-context-length 8 \
  --memory-size 32 \
  --decoding-method fast_beam_search \
  --use-averaged-model True \
  --beam 4 \
  --max-contexts 4 \
  --max-states 8

The simulated streaming decoding command using modified beam search is:

./conv_emformer_transducer_stateless2/decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless2/exp \
  --max-duration 300 \
  --num-encoder-layers 12 \
  --chunk-length 32 \
  --cnn-module-kernel 31 \
  --left-context-length 32 \
  --right-context-length 8 \
  --memory-size 32 \
  --decoding-method modified_beam_search \
  --use-averaged-model True \
  --beam-size 4

The streaming decoding command using greedy search is:

./conv_emformer_transducer_stateless2/streaming_decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless2/exp \
  --num-decode-streams 2000 \
  --num-encoder-layers 12 \
  --chunk-length 32 \
  --cnn-module-kernel 31 \
  --left-context-length 32 \
  --right-context-length 8 \
  --memory-size 32 \
  --decoding-method greedy_search \
  --use-averaged-model True

The streaming decoding command using fast beam search is:

./conv_emformer_transducer_stateless2/streaming_decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless2/exp \
  --num-decode-streams 2000 \
  --num-encoder-layers 12 \
  --chunk-length 32 \
  --cnn-module-kernel 31 \
  --left-context-length 32 \
  --right-context-length 8 \
  --memory-size 32 \
  --decoding-method fast_beam_search \
  --use-averaged-model True \
  --beam 4 \
  --max-contexts 4 \
  --max-states 8

The streaming decoding command using modified beam search is:

./conv_emformer_transducer_stateless2/streaming_decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless2/exp \
  --num-decode-streams 2000 \
  --num-encoder-layers 12 \
  --chunk-length 32 \
  --cnn-module-kernel 31 \
  --left-context-length 32 \
  --right-context-length 8 \
  --memory-size 32 \
  --decoding-method modified_beam_search \
  --use-averaged-model True \
  --beam-size 4

Pretrained models, training logs, decoding logs, and decoding results are available at https://huggingface.co/Zengwei/icefall-asr-librispeech-conv-emformer-transducer-stateless2-2022-07-05

With higher latency setup, training on full librispeech

In this model, the lengths of chunk and right context are 64 frames (i.e., 0.64s) and 16 frames (i.e., 0.16s), respectively.

The WERs are:

	test-clean	test-other	comment	decoding mode
greedy search (max sym per frame 1)	3.3	8.71	--epoch 30 --avg 10	simulated streaming
greedy search (max sym per frame 1)	3.35	8.65	--epoch 30 --avg 10	streaming
fast beam search	3.27	8.58	--epoch 30 --avg 10	simulated streaming
fast beam search	3.31	8.48	--epoch 30 --avg 10	streaming
modified beam search	3.26	8.56	--epoch 30 --avg 10	simulated streaming
modified beam search	3.29	8.47	--epoch 30 --avg 10	streaming

The training command is:

./conv_emformer_transducer_stateless2/train.py \
  --world-size 4 \
  --num-epochs 30 \
  --start-epoch 1 \
  --exp-dir conv_emformer_transducer_stateless2/exp \
  --full-libri 1 \
  --max-duration 280 \
  --master-port 12321 \
  --num-encoder-layers 12 \
  --chunk-length 64 \
  --cnn-module-kernel 31 \
  --left-context-length 64 \
  --right-context-length 16 \
  --memory-size 32

The tensorboard log can be found at https://tensorboard.dev/experiment/eRx6XwbOQhGlywgD8lWBjw/

The simulated streaming decoding command using greedy search is:

./conv_emformer_transducer_stateless2/decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless2/exp \
  --max-duration 300 \
  --num-encoder-layers 12 \
  --chunk-length 64 \
  --cnn-module-kernel 31 \
  --left-context-length 64 \
  --right-context-length 16 \
  --memory-size 32 \
  --decoding-method greedy_search \
  --use-averaged-model True

The simulated streaming decoding command using fast beam search is:

./conv_emformer_transducer_stateless2/decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless2/exp \
  --max-duration 300 \
  --num-encoder-layers 12 \
  --chunk-length 64 \
  --cnn-module-kernel 31 \
  --left-context-length 64 \
  --right-context-length 16 \
  --memory-size 32 \
  --decoding-method fast_beam_search \
  --use-averaged-model True \
  --beam 4 \
  --max-contexts 4 \
  --max-states 8

The simulated streaming decoding command using modified beam search is:

./conv_emformer_transducer_stateless2/decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless2/exp \
  --max-duration 300 \
  --num-encoder-layers 12 \
  --chunk-length 64 \
  --cnn-module-kernel 31 \
  --left-context-length 64 \
  --right-context-length 16 \
  --memory-size 32 \
  --decoding-method modified_beam_search \
  --use-averaged-model True \
  --beam-size 4

The streaming decoding command using greedy search is:

./conv_emformer_transducer_stateless2/streaming_decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless2/exp \
  --num-decode-streams 2000 \
  --num-encoder-layers 12 \
  --chunk-length 64 \
  --cnn-module-kernel 31 \
  --left-context-length 64 \
  --right-context-length 16 \
  --memory-size 32 \
  --decoding-method greedy_search \
  --use-averaged-model True

The streaming decoding command using fast beam search is:

./conv_emformer_transducer_stateless2/streaming_decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless2/exp \
  --num-decode-streams 2000 \
  --num-encoder-layers 12 \
  --chunk-length 64 \
  --cnn-module-kernel 31 \
  --left-context-length 64 \
  --right-context-length 16 \
  --memory-size 32 \
  --decoding-method fast_beam_search \
  --use-averaged-model True \
  --beam 4 \
  --max-contexts 4 \
  --max-states 8

The streaming decoding command using modified beam search is:

./conv_emformer_transducer_stateless2/streaming_decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless2/exp \
  --num-decode-streams 2000 \
  --num-encoder-layers 12 \
  --chunk-length 64 \
  --cnn-module-kernel 31 \
  --left-context-length 64 \
  --right-context-length 16 \
  --memory-size 32 \
  --decoding-method modified_beam_search \
  --use-averaged-model True \
  --beam-size 4

Pretrained models, training logs, decoding logs, and decoding results are available at https://huggingface.co/Zengwei/icefall-asr-librispeech-conv-emformer-transducer-stateless2-larger-latency-2022-07-06

LibriSpeech BPE training results (Pruned Stateless Streaming Conformer RNN-T)

pruned_transducer_stateless

See #380 for more details.

Training on full librispeech

The WERs are (the number in the table formatted as test-clean & test-other):

We only trained 25 epochs for saving time, if you want to get better results you can train more epochs.

decoding method	left context	chunk size = 2	chunk size = 4	chunk size = 8	chunk size = 16
greedy search	32	4.74 & 11.38	4.57 & 10.86	4.18 & 10.37	3.87 & 9.85
greedy search	64	4.74 & 11.25	4.48 & 10.72	4.1 & 10.24	3.85 & 9.73
fast beam search	32	4.75 & 11.1	4.48 & 10.65	4.12 & 10.18	3.95 & 9.67
fast beam search	64	4.7 & 11	4.37 & 10.49	4.07 & 10.04	3.89 & 9.53
modified beam search	32	4.64 & 10.94	4.38 & 10.51	4.11 & 10.14	3.87 & 9.61
modified beam search	64	4.59 & 10.81	4.29 & 10.39	4.02 & 10.02	3.84 & 9.43

NOTE: The WERs in table above were decoded with simulate streaming method (i.e. using masking strategy), see commands below. We also have real streaming decoding script which should produce almost the same results. We tried adding right context in the real streaming decoding, but it seemed not to benefit the performance for all the models, the reasons might be the training and decoding mismatching.

The training command is:

./pruned_transducer_stateless/train.py \
  --exp-dir pruned_transducer_stateless/exp \
  --full-libri 1 \
  --dynamic-chunk-training 1 \
  --causal-convolution 1 \
  --short-chunk-size 20 \
  --num-left-chunks 4 \
  --max-duration 300 \
  --world-size 4 \
  --start-epoch 0 \
  --num-epochs 25

You can find the tensorboard log here https://tensorboard.dev/experiment/ofxRakE6R7WHB1AoB8Bweg/

The decoding command is:

decoding_method="greedy_search"  # "fast_beam_search", "modified_beam_search"

for chunk in 2 4 8 16; do
  for left in 32 64; do
    ./pruned_transducer_stateless/decode.py \
            --simulate-streaming 1 \
            --decode-chunk-size ${chunk} \
            --left-context ${left} \
            --causal-convolution 1 \
            --epoch 24 \
            --avg 10 \
            --exp-dir ./pruned_transducer_stateless/exp \
            --max-sym-per-frame 1 \
            --max-duration 1000 \
            --decoding-method ${decoding_method}
  done
done

Pre-trained models, training and decoding logs, and decoding results are available at https://huggingface.co/pkufool/icefall_librispeech_streaming_pruned_transducer_stateless_20220625

pruned_transducer_stateless2

See #380 for more details.

Training on full librispeech

The WERs are (the number in the table formatted as test-clean & test-other):

We only trained 25 epochs for saving time, if you want to get better results you can train more epochs.

decoding method	left context	chunk size = 2	chunk size = 4	chunk size = 8	chunk size = 16
greedy search	32	4.2 & 10.64	3.97 & 10.03	3.83 & 9.58	3.7 & 9.11
greedy search	64	4.16 & 10.5	3.93 & 9.99	3.73 & 9.45	3.63 & 9.04
fast beam search	32	4.13 & 10.3	3.93 & 9.82	3.8 & 9.35	3.62 & 8.93
fast beam search	64	4.13 & 10.22	3.89 & 9.68	3.73 & 9.27	3.52 & 8.82
modified beam search	32	4.02 & 10.22	3.9 & 9.71	3.74 & 9.33	3.59 & 8.87
modified beam search	64	4.05 & 10.08	3.81 & 9.67	3.68 & 9.21	3.56 & 8.77

NOTE: The WERs in table above were decoded with simulate streaming method (i.e. using masking strategy), see commands below. We also have real streaming decoding script which should produce almost the same results. We tried adding right context in the real streaming decoding, but it seemed not to benefit the performance for all the models, the reasons might be the training and decoding mismatching.

The training command is:

./pruned_transducer_stateless2/train.py \
  --exp-dir pruned_transducer_stateless2/exp \
  --full-libri 1 \
  --dynamic-chunk-training 1 \
  --causal-convolution 1 \
  --short-chunk-size 20 \
  --num-left-chunks 4 \
  --max-duration 300 \
  --world-size 4 \
  --start-epoch 0 \
  --num-epochs 25

You can find the tensorboard log here https://tensorboard.dev/experiment/hbltNS5TQ1Kiw0D1vcoakw/

The decoding command is:

decoding_method="greedy_search"  # "fast_beam_search", "modified_beam_search"

for chunk in 2 4 8 16; do
  for left in 32 64; do
    ./pruned_transducer_stateless2/decode.py \
            --simulate-streaming 1 \
            --decode-chunk-size ${chunk} \
            --left-context ${left} \
            --causal-convolution 1 \
            --epoch 24 \
            --avg 10 \
            --exp-dir ./pruned_transducer_stateless2/exp \
            --max-sym-per-frame 1 \
            --max-duration 1000 \
            --decoding-method ${decoding_method}
  done
done

Pre-trained models, training and decoding logs, and decoding results are available at https://huggingface.co/pkufool/icefall_librispeech_streaming_pruned_transducer_stateless2_20220625

pruned_transducer_stateless3

See #380 for more details.

Training on full librispeech (Use giga_prob = 0.5)

The WERs are (the number in the table formatted as test-clean & test-other):

decoding method	left context	chunk size = 2	chunk size = 4	chunk size = 8	chunk size = 16
greedy search	32	3.7 & 9.53	3.45 & 8.88	3.28 & 8.45	3.13 & 7.93
greedy search	64	3.69 & 9.36	3.39 & 8.68	3.28 & 8.19	3.08 & 7.83
fast beam search	32	3.71 & 9.18	3.36 & 8.65	3.23 & 8.23	3.17 & 7.78
fast beam search	64	3.61 & 9.03	3.46 & 8.43	3.2 & 8.0	3.11 & 7.63
modified beam search	32	3.56 & 9.08	3.34 & 8.58	3.21 & 8.14	3.06 & 7.73
modified beam search	64	3.55 & 8.86	3.29 & 8.34	3.16 & 8.01	3.05 & 7.57

NOTE: The WERs in table above were decoded with simulate streaming method (i.e. using masking strategy), see commands below. We also have real streaming decoding script which should produce almost the same results. We tried adding right context in the real streaming decoding, but it seemed not to benefit the performance for all the models, the reasons might be the training and decoding mismatching.

The training command is (Note: this model was trained with mix-precision training):

./pruned_transducer_stateless3/train.py \
  --exp-dir pruned_transducer_stateless3/exp \
  --full-libri 1 \
  --dynamic-chunk-training 1 \
  --causal-convolution 1 \
  --short-chunk-size 32 \
  --num-left-chunks 4 \
  --max-duration 300 \
  --world-size 4 \
  --use-fp16 1 \
  --start-epoch 0 \
  --num-epochs 37 \
  --num-workers 2 \
  --giga-prob 0.5

You can find the tensorboard log here https://tensorboard.dev/experiment/vL7dWVZqTYaSeoOED4rtow/

The decoding command is:

decoding_method="greedy_search"  # "fast_beam_search", "modified_beam_search"

for chunk in 2 4 8 16; do
  for left in 32 64; do
    ./pruned_transducer_stateless3/decode.py \
            --simulate-streaming 1 \
            --decode-chunk-size ${chunk} \
            --left-context ${left} \
            --causal-convolution 1 \
            --epoch 36 \
            --avg 8 \
            --exp-dir ./pruned_transducer_stateless3/exp \
            --max-sym-per-frame 1 \
            --max-duration 1000 \
            --decoding-method ${decoding_method}
  done
done

Pre-trained models, training and decoding logs, and decoding results are available at https://huggingface.co/pkufool/icefall_librispeech_streaming_pruned_transducer_stateless3_giga_0.5_20220625

Training on full librispeech (Use giga_prob = 0.9)

The WERs are (the number in the table formatted as test-clean & test-other):

decoding method	left context	chunk size = 2	chunk size = 4	chunk size = 8	chunk size = 16
greedy search	32	3.25 & 8.2	3.07 & 7.67	2.91 & 7.28	2.8 & 6.89
greedy search	64	3.22 & 8.12	3.05 & 7.59	2.91 & 7.07	2.78 & 6.81
fast beam search	32	3.26 & 8.2	3.06 & 7.56	2.98 & 7.08	2.77 & 6.75
fast beam search	64	3.24 & 8.09	3.06 & 7.43	2.88 & 7.03	2.73 & 6.68
modified beam search	32	3.13 & 7.91	2.99 & 7.45	2.83 & 6.98	2.68 & 6.75
modified beam search	64	3.08 & 7.8	2.97 & 7.37	2.81 & 6.82	2.66 & 6.67

NOTE: The WERs in table above were decoded with simulate streaming method (i.e. using masking strategy), see commands below. We also have real streaming decoding script which should produce almost the same results. We tried adding right context in the real streaming decoding, but it seemed not to benefit the performance for all the models, the reasons might be the training and decoding mismatching.

The training command is:

./pruned_transducer_stateless3/train.py \
  --exp-dir pruned_transducer_stateless3/exp \
  --full-libri 1 \
  --dynamic-chunk-training 1 \
  --causal-convolution 1 \
  --short-chunk-size 25 \
  --num-left-chunks 8 \
  --max-duration 300 \
  --world-size 8 \
  --start-epoch 0 \
  --num-epochs 26 \
  --num-workers 2 \
  --giga-prob 0.9

You can find the tensorboard log here https://tensorboard.dev/experiment/WBGBDzt7SByRnvCBEfQpGQ/

The decoding command is:

decoding_method="greedy_search"  # "fast_beam_search", "modified_beam_search"

for chunk in 2 4 8 16; do
  for left in 32 64; do
    ./pruned_transducer_stateless3/decode.py \
            --simulate-streaming 1 \
            --decode-chunk-size ${chunk} \
            --left-context ${left} \
            --causal-convolution 1 \
            --epoch 25 \
            --avg 12 \
            --exp-dir ./pruned_transducer_stateless3/exp \
            --max-sym-per-frame 1 \
            --max-duration 1000 \
            --decoding-method ${decoding_method}
  done
done

Pre-trained models, training and decoding logs, and decoding results are available at https://huggingface.co/pkufool/icefall_librispeech_streaming_pruned_transducer_stateless3_giga_0.9_20220625

pruned_transducer_stateless4

See #380 for more details.

Training on full librispeech

The WERs are (the number in the table formatted as test-clean & test-other):

We only trained 25 epochs for saving time, if you want to get better results you can train more epochs.

decoding method	left context	chunk size = 2	chunk size = 4	chunk size = 8	chunk size = 16
greedy search	32	3.96 & 10.45	3.73 & 9.97	3.54 & 9.56	3.45 & 9.08
greedy search	64	3.9 & 10.34	3.7 & 9.9	3.53 & 9.41	3.39 & 9.03
fast beam search	32	3.9 & 10.09	3.69 & 9.65	3.58 & 9.28	3.46 & 8.91
fast beam search	64	3.82 & 10.03	3.67 & 9.56	3.51 & 9.18	3.43 & 8.78
modified beam search	32	3.78 & 10.0	3.63 & 9.54	3.43 & 9.29	3.39 & 8.84
modified beam search	64	3.76 & 9.95	3.54 & 9.48	3.4 & 9.13	3.33 & 8.74

NOTE: The WERs in table above were decoded with simulate streaming method (i.e. using masking strategy), see commands below. We also have real streaming decoding script which should produce almost the same results. We tried adding right context in the real streaming decoding, but it seemed not to benefit the performance for all the models, the reasons might be the training and decoding mismatching.

The training command is:

./pruned_transducer_stateless4/train.py \
  --exp-dir pruned_transducer_stateless4/exp \
  --full-libri 1 \
  --dynamic-chunk-training 1 \
  --causal-convolution 1 \
  --short-chunk-size 20 \
  --num-left-chunks 4 \
  --max-duration 300 \
  --world-size 4 \
  --start-epoch 1 \
  --num-epochs 25

You can find the tensorboard log here https://tensorboard.dev/experiment/97VKXf80Ru61CnP2ALWZZg/

The decoding command is:

decoding_method="greedy_search"  # "fast_beam_search", "modified_beam_search"

for chunk in 2 4 8 16; do
  for left in 32 64; do
    ./pruned_transducer_stateless4/decode.py \
            --simulate-streaming 1 \
            --decode-chunk-size ${chunk} \
            --left-context ${left} \
            --causal-convolution 1 \
            --epoch 25 \
            --avg 3 \
            --exp-dir ./pruned_transducer_stateless4/exp \
            --max-sym-per-frame 1 \
            --max-duration 1000 \
            --decoding-method ${decoding_method}
  done
done

Pre-trained models, training and decoding logs, and decoding results are available at https://huggingface.co/pkufool/icefall_librispeech_streaming_pruned_transducer_stateless4_20220625

pruned_transducer_stateless5

See #454 for more details.

Training on full librispeech

The WERs are (the number in the table formatted as test-clean & test-other):

We only trained 25 epochs for saving time, if you want to get better results you can train more epochs.

decoding method	left context	chunk size = 2	chunk size = 4	chunk size = 8	chunk size = 16
greedy search	32	3.93 & 9.88	3.64 & 9.43	3.51 & 8.92	3.26 & 8.37
greedy search	64	4.84 & 9.81	3.59 & 9.27	3.44 & 8.83	3.23 & 8.33
fast beam search	32	3.86 & 9.77	3.67 & 9.3	3.5 & 8.83	3.27 & 8.33
fast beam search	64	3.79 & 9.68	3.57 & 9.21	3.41 & 8.72	3.25 & 8.27
modified beam search	32	3.84 & 9.71	3.66 & 9.38	3.47 & 8.86	3.26 & 8.42
modified beam search	64	3.81 & 9.59	3.58 & 9.2	3.44 & 8.74	3.23 & 8.35

NOTE: The WERs in table above were decoded with simulate streaming method (i.e. using masking strategy), see commands below. We also have real streaming decoding script which should produce almost the same results. We tried adding right context in the real streaming decoding, but it seemed not to benefit the performance for all the models, the reasons might be the training and decoding mismatching.

The training command is:

./pruned_transducer_stateless5/train.py \
  --exp-dir pruned_transducer_stateless5/exp \
  --num-encoder-layers 18 \
  --dim-feedforward 2048 \
  --nhead 8 \
  --encoder-dim 512 \
  --decoder-dim 512 \
  --joiner-dim 512 \
  --full-libri 1 \
  --dynamic-chunk-training 1 \
  --causal-convolution 1 \
  --short-chunk-size 20 \
  --num-left-chunks 4 \
  --max-duration 300 \
  --world-size 4 \
  --start-epoch 1 \
  --num-epochs 25

You can find the tensorboard log here https://tensorboard.dev/experiment/rO04h6vjTLyw0qSxjp4m4Q

The decoding command is:

decoding_method="greedy_search"  # "fast_beam_search", "modified_beam_search"

for chunk in 2 4 8 16; do
  for left in 32 64; do
    ./pruned_transducer_stateless5/decode.py \
            --num-encoder-layers 18 \
            --dim-feedforward 2048 \
            --nhead 8 \
            --encoder-dim 512 \
            --decoder-dim 512 \
            --joiner-dim 512 \
            --simulate-streaming 1 \
            --decode-chunk-size ${chunk} \
            --left-context ${left} \
            --causal-convolution 1 \
            --epoch 25 \
            --avg 3 \
            --exp-dir ./pruned_transducer_stateless5/exp \
            --max-sym-per-frame 1 \
            --max-duration 1000 \
            --decoding-method ${decoding_method}
  done
done

Pre-trained models, training and decoding logs, and decoding results are available at https://huggingface.co/pkufool/icefall_librispeech_streaming_pruned_transducer_stateless5_20220729

LibriSpeech BPE training results (Pruned Stateless Conv-Emformer RNN-T)

conv_emformer_transducer_stateless

It implements Emformer augmented with convolution module for streaming ASR. It is modified from torchaudio.

See #389 for more details.

Training on full librispeech

In this model, the lengths of chunk and right context are 32 frames (i.e., 0.32s) and 8 frames (i.e., 0.08s), respectively.

The WERs are:

	test-clean	test-other	comment	decoding mode
greedy search (max sym per frame 1)	3.63	9.61	--epoch 30 --avg 10	simulated streaming
greedy search (max sym per frame 1)	3.64	9.65	--epoch 30 --avg 10	streaming
fast beam search	3.61	9.4	--epoch 30 --avg 10	simulated streaming
fast beam search	3.58	9.5	--epoch 30 --avg 10	streaming
modified beam search	3.56	9.41	--epoch 30 --avg 10	simulated streaming
modified beam search	3.54	9.46	--epoch 30 --avg 10	streaming

The training command is:

./conv_emformer_transducer_stateless/train.py \
  --world-size 6 \
  --num-epochs 30 \
  --start-epoch 1 \
  --exp-dir conv_emformer_transducer_stateless/exp \
  --full-libri 1 \
  --max-duration 300 \
  --master-port 12321 \
  --num-encoder-layers 12 \
  --chunk-length 32 \
  --cnn-module-kernel 31 \
  --left-context-length 32 \
  --right-context-length 8 \
  --memory-size 32

The tensorboard log can be found at https://tensorboard.dev/experiment/4em2FLsxRwGhmoCRQUEoDw/

The simulated streaming decoding command using greedy search is:

./conv_emformer_transducer_stateless/decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless/exp \
  --max-duration 300 \
  --num-encoder-layers 12 \
  --chunk-length 32 \
  --cnn-module-kernel 31 \
  --left-context-length 32 \
  --right-context-length 8 \
  --memory-size 32 \
  --decoding-method greedy_search \
  --use-averaged-model True

The simulated streaming decoding command using fast beam search is:

./conv_emformer_transducer_stateless/decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless/exp \
  --max-duration 300 \
  --num-encoder-layers 12 \
  --chunk-length 32 \
  --cnn-module-kernel 31 \
  --left-context-length 32 \
  --right-context-length 8 \
  --memory-size 32 \
  --decoding-method fast_beam_search \
  --use-averaged-model True \
  --beam 4 \
  --max-contexts 4 \
  --max-states 8

The simulated streaming decoding command using modified beam search is:

./conv_emformer_transducer_stateless/decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless/exp \
  --max-duration 300 \
  --num-encoder-layers 12 \
  --chunk-length 32 \
  --cnn-module-kernel 31 \
  --left-context-length 32 \
  --right-context-length 8 \
  --memory-size 32 \
  --decoding-method modified_beam_search \
  --use-averaged-model True \
  --beam-size 4

The streaming decoding command using greedy search is:

./conv_emformer_transducer_stateless/streaming_decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless/exp \
  --num-decode-streams 2000 \
  --num-encoder-layers 12 \
  --chunk-length 32 \
  --cnn-module-kernel 31 \
  --left-context-length 32 \
  --right-context-length 8 \
  --memory-size 32 \
  --decoding-method greedy_search \
  --use-averaged-model True

The streaming decoding command using fast beam search is:

./conv_emformer_transducer_stateless/streaming_decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless/exp \
  --num-decode-streams 2000 \
  --num-encoder-layers 12 \
  --chunk-length 32 \
  --cnn-module-kernel 31 \
  --left-context-length 32 \
  --right-context-length 8 \
  --memory-size 32 \
  --decoding-method fast_beam_search \
  --use-averaged-model True \
  --beam 4 \
  --max-contexts 4 \
  --max-states 8

The streaming decoding command using modified beam search is:

./conv_emformer_transducer_stateless/streaming_decode.py \
  --epoch 30 \
  --avg 10 \
  --exp-dir conv_emformer_transducer_stateless/exp \
  --num-decode-streams 2000 \
  --num-encoder-layers 12 \
  --chunk-length 32 \
  --cnn-module-kernel 31 \
  --left-context-length 32 \
  --right-context-length 8 \
  --memory-size 32 \
  --decoding-method modified_beam_search \
  --use-averaged-model True \
  --beam-size 4

Pretrained models, training logs, decoding logs, and decoding results are available at https://huggingface.co/Zengwei/icefall-asr-librispeech-conv-emformer-transducer-stateless-2022-06-11

LibriSpeech BPE training results (Pruned Stateless Emformer RNN-T)

pruned_stateless_emformer_rnnt2

Use #390.

Use Emformer from torchaudio for streaming ASR. The Emformer model is imported from torchaudio without modifications.

You can use https://github.com/k2-fsa/sherpa to deploy it.

	test-clean	test-other	comment
greedy search (max sym per frame 1)	4.28	11.42	--epoch 39 --avg 6 --max-duration 600
modified beam search	4.22	11.16	--epoch 39 --avg 6 --max-duration 600
fast beam search	4.29	11.26	--epoch 39 --avg 6 --max-duration 600

The training commands are:

export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"

./pruned_stateless_emformer_rnnt2/train.py \
  --world-size 8 \
  --num-epochs 40 \
  --start-epoch 1 \
  --exp-dir pruned_stateless_emformer_rnnt2/exp-full \
  --full-libri 1 \
  --use-fp16 0 \
  --max-duration 200 \
  --prune-range 5 \
  --lm-scale 0.25 \
  --master-port 12358 \
  --num-encoder-layers 18 \
  --left-context-length 128 \
  --segment-length 8 \
  --right-context-length 4

The tensorboard log can be found at https://tensorboard.dev/experiment/ZyiqhAhmRjmr49xml4ofLw/

The decoding commands are:

for m in greedy_search fast_beam_search modified_beam_search; do
  for epoch in 39; do
    for avg in 6; do
      ./pruned_stateless_emformer_rnnt2/decode.py \
        --epoch $epoch \
        --avg $avg \
        --use-averaged-model 1 \
        --exp-dir pruned_stateless_emformer_rnnt2/exp-full \
        --max-duration 50 \
        --decoding-method $m \
        --num-encoder-layers 18 \
        --left-context-length 128 \
        --segment-length 8 \
        --right-context-length 4
    done
  done
done

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-stateless-emformer-rnnt2-2022-06-01

LibriSpeech BPE training results (Pruned Stateless Transducer 5)

pruned_transducer_stateless5

Same as Pruned Stateless Transducer 2 but with more layers.

See #330

Note that models in pruned_transducer_stateless and pruned_transducer_stateless2 have about 80 M parameters.

The notations large and medium below are from the Conformer paper, where the large model has about 118 M parameters and the medium model has 30.8 M parameters.

Large

Number of model parameters 118129516 (i.e, 118.13 M).

	test-clean	test-other	comment
greedy search (max sym per frame 1)	2.43	5.72	--epoch 30 --avg 10 --max-duration 600
modified beam search	2.43	5.69	--epoch 30 --avg 10 --max-duration 600
fast beam search	2.43	5.67	--epoch 30 --avg 10 --max-duration 600

The training commands are:

export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"

./pruned_transducer_stateless5/train.py \
  --world-size 8 \
  --num-epochs 30 \
  --start-epoch 1 \
  --full-libri 1 \
  --exp-dir pruned_transducer_stateless5/exp-L \
  --max-duration 300 \
  --use-fp16 0 \
  --num-encoder-layers 18 \
  --dim-feedforward 2048 \
  --nhead 8 \
  --encoder-dim 512 \
  --decoder-dim 512 \
  --joiner-dim 512

The tensorboard log can be found at https://tensorboard.dev/experiment/aWzDj5swSE2VmcOYgoe3vQ

The decoding commands are:

for method in greedy_search modified_beam_search fast_beam_search; do
  ./pruned_transducer_stateless5/decode.py \
    --epoch 30 \
    --avg 10 \
    --exp-dir ./pruned_transducer_stateless5/exp-L \
    --max-duration 600 \
    --decoding-method $method \
    --max-sym-per-frame 1 \
    --num-encoder-layers 18 \
    --dim-feedforward 2048 \
    --nhead 8 \
    --encoder-dim 512 \
    --decoder-dim 512 \
    --joiner-dim 512 \
    --use-averaged-model True
done

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless5-2022-07-07

Medium

Number of model parameters 30896748 (i.e, 30.9 M).

	test-clean	test-other	comment
greedy search (max sym per frame 1)	2.87	6.92	--epoch 30 --avg 10 --max-duration 600
modified beam search	2.83	6.75	--epoch 30 --avg 10 --max-duration 600
fast beam search	2.81	6.76	--epoch 30 --avg 10 --max-duration 600

The training commands are:

export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"

./pruned_transducer_stateless5/train.py \
  --world-size 8 \
  --num-epochs 30 \
  --start-epoch 1 \
  --full-libri 1 \
  --exp-dir pruned_transducer_stateless5/exp-M \
  --max-duration 300 \
  --use-fp16 0 \
  --num-encoder-layers 18 \
  --dim-feedforward 1024 \
  --nhead 4 \
  --encoder-dim 256 \
  --decoder-dim 512 \
  --joiner-dim 512

The tensorboard log can be found at https://tensorboard.dev/experiment/04xtWUKPRmebSnpzN1GMHQ

The decoding commands are:

for method in greedy_search modified_beam_search fast_beam_search; do
  ./pruned_transducer_stateless5/decode.py \
    --epoch 30 \
    --avg 10 \
    --exp-dir ./pruned_transducer_stateless5/exp-M \
    --max-duration 600 \
    --decoding-method $method \
    --max-sym-per-frame 1 \
    --num-encoder-layers 18 \
    --dim-feedforward 1024 \
    --nhead 4 \
    --encoder-dim 256 \
    --decoder-dim 512 \
    --joiner-dim 512 \
    --use-averaged-model True
done

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless5-M-2022-07-07

Baseline-2

It has 87.8 M parameters. Compared to the model in pruned_transducer_stateless2, its has more layers (24 v.s 12) but a narrower model (1536 feedforward dim and 384 encoder dim vs 2048 feed forward dim and 512 encoder dim).

	test-clean	test-other	comment
greedy search (max sym per frame 1)	2.54	5.72	--epoch 30 --avg 10 --max-duration 600
modified beam search	2.47	5.71	--epoch 30 --avg 10 --max-duration 600
modified beam search + RNNLM shallow fusion	2.27	5.24	--epoch 30 --avg 10 --max-duration 600
modified beam search + RNNLM + LODR	2.23	5.17	--epoch 30 --avg 10 --max-duration 600
modified beam search + TransformerLM shallow fusion	2.27	5.26	--epoch 30 --avg 10 --max-duration 600
modified beam search + TransformerLM + LODR	2.22	5.11	--epoch 30 --avg 10 --max-duration 600
fast beam search	2.5	5.72	--epoch 30 --avg 10 --max-duration 600

export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"

./pruned_transducer_stateless5/train.py \
  --world-size 8 \
  --num-epochs 30 \
  --start-epoch 1 \
  --full-libri 1 \
  --exp-dir pruned_transducer_stateless5/exp-B \
  --max-duration 300 \
  --use-fp16 0 \
  --num-encoder-layers 24 \
  --dim-feedforward 1536 \
  --nhead 8 \
  --encoder-dim 384 \
  --decoder-dim 512 \
  --joiner-dim 512

The tensorboard log can be found at https://tensorboard.dev/experiment/foVHNyqiRi2LhybmRUOAyg

The decoding commands are:

for method in greedy_search modified_beam_search fast_beam_search; do
  ./pruned_transducer_stateless5/decode.py \
    --epoch 30 \
    --avg 10 \
    --exp-dir ./pruned_transducer_stateless5/exp-B \
    --max-duration 600 \
    --decoding-method $method \
    --max-sym-per-frame 1 \
    --num-encoder-layers 24 \
    --dim-feedforward 1536 \
    --nhead 8 \
    --encoder-dim 384 \
    --decoder-dim 512 \
    --joiner-dim 512 \
    --use-averaged-model True
done

To decode with RNNLM shallow fusion, use the following decoding command. A well-trained RNNLM can be found here: https://huggingface.co/ezerhouni/icefall-librispeech-rnn-lm/tree/main

for method in greedy_search modified_beam_search fast_beam_search; do
  ./pruned_transducer_stateless5/decode.py \
    --epoch 30 \
    --avg 10 \
    --exp-dir ./pruned_transducer_stateless5/exp-B \
    --max-duration 600 \
    --decoding-method modified_beam_search_rnnlm_shallow_fusion \
    --max-sym-per-frame 1 \
    --num-encoder-layers 24 \
    --dim-feedforward 1536 \
    --nhead 8 \
    --encoder-dim 384 \
    --decoder-dim 512 \
    --joiner-dim 512 \
    --use-averaged-model True
    --beam 4 \
    --max-contexts 4 \
    --rnn-lm-scale 0.4 \
    --rnn-lm-exp-dir /path/to/RNNLM/exp \
    --rnn-lm-epoch 99 \
    --rnn-lm-avg 1 \
    --rnn-lm-num-layers 3 \
    --rnn-lm-tie-weights 1
done

You can find a pretrained model, training logs, decoding logs, and decoding results at: https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless5-B-2022-07-07

LibriSpeech BPE training results (Pruned Stateless Transducer 4)

pruned_transducer_stateless4

This version saves averaged model during training, and decodes with averaged model.

See #337 for details about the idea of model averaging.

Training on full librispeech

See #344

Using commit ec0b0e92297cc03fdb09f48cd235e84d2c04156b.

The WERs are:

	test-clean	test-other	comment
greedy search (max sym per frame 1)	2.75	6.74	--epoch 30 --avg 6 --use-averaged-model False
greedy search (max sym per frame 1)	2.69	6.64	--epoch 30 --avg 6 --use-averaged-model True
fast beam search	2.72	6.67	--epoch 30 --avg 6 --use-averaged-model False
fast beam search	2.66	6.6	--epoch 30 --avg 6 --use-averaged-model True
modified beam search	2.67	6.68	--epoch 30 --avg 6 --use-averaged-model False
modified beam search	2.62	6.57	--epoch 30 --avg 6 --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

The tensorboard log can be found at https://tensorboard.dev/experiment/QOGSPBgsR8KzcRMmie9JGw/

The decoding command using greedy search is:

./pruned_transducer_stateless4/decode.py \
  --epoch 30 \
  --avg 6 \
  --exp-dir pruned_transducer_stateless4/exp \
  --max-duration 300 \
  --decoding-method greedy_search \
  --use-averaged-model True

The decoding command using fast beam search is:

./pruned_transducer_stateless4/decode.py \
  --epoch 30 \
  --avg 6 \
  --exp-dir pruned_transducer_stateless4/exp \
  --max-duration 300 \
  --decoding-method fast_beam_search \
  --use-averaged-model True \
  --beam 4 \
  --max-contexts 4 \
  --max-states 8

The decoding command using modified beam search is:

./pruned_transducer_stateless4/decode.py \
  --epoch 30 \
  --avg 6 \
  --exp-dir pruned_transducer_stateless4/exp \
  --max-duration 300 \
  --decoding-method modified_beam_search \
  --use-averaged-model True \
  --beam-size 4

Pretrained models, training logs, decoding logs, and decoding results are available at https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless4-2022-06-03

Training on train-clean-100

See #344

Using commit ec0b0e92297cc03fdb09f48cd235e84d2c04156b.

The WERs are:

	test-clean	test-other	comment
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	6.82	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	6.74	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

The tensorboard log can be found at https://tensorboard.dev/experiment/YVYHq1irQS69s9bW1vQ06Q/

LibriSpeech BPE training results (Pruned Stateless Transducer 3, 2022-04-29)

pruned_transducer_stateless3

Same as Pruned Stateless Transducer 2 but using the XL subset from GigaSpeech as extra training data.

During training, it selects either a batch from GigaSpeech with prob giga_prob or a batch from LibriSpeech with prob 1 - giga_prob. All utterances within a batch come from the same dataset.

Using commit ac84220de91dee10c00e8f4223287f937b1930b6.

See #312.

The WERs are:

	test-clean	test-other	comment
greedy search (max sym per frame 1)	2.21	5.09	--epoch 27 --avg 2 --max-duration 600
greedy search (max sym per frame 1)	2.25	5.02	--epoch 27 --avg 12 --max-duration 600
modified beam search	2.19	5.03	--epoch 25 --avg 6 --max-duration 600
modified beam search	2.23	4.94	--epoch 27 --avg 10 --max-duration 600
beam search	2.16	4.95	--epoch 25 --avg 7 --max-duration 600
fast beam search	2.21	4.96	--epoch 27 --avg 10 --max-duration 600
fast beam search	2.19	4.97	--epoch 27 --avg 12 --max-duration 600

The training commands are:

./prepare.sh
./prepare_giga_speech.sh

export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"

./pruned_transducer_stateless3/train.py \
  --world-size 8 \
  --num-epochs 30 \
  --start-epoch 0 \
  --full-libri 1 \
  --exp-dir pruned_transducer_stateless3/exp \
  --max-duration 300 \
  --use-fp16 1 \
  --lr-epochs 4 \
  --num-workers 2 \
  --giga-prob 0.8

The tensorboard log can be found at https://tensorboard.dev/experiment/gaD34WeYSMCOkzoo3dZXGg/ (Note: The training process is killed manually after saving epoch-28.pt.)

Pretrained models, training logs, decoding logs, and decoding results are available at https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-04-29

The decoding commands are:

# greedy search
./pruned_transducer_stateless3/decode.py \
    --epoch 27 \
    --avg 2 \
    --exp-dir ./pruned_transducer_stateless3/exp \
    --max-duration 600 \
    --decoding-method greedy_search \
    --max-sym-per-frame 1

# modified beam search
./pruned_transducer_stateless3/decode.py \
    --epoch 25 \
    --avg 6 \
    --exp-dir ./pruned_transducer_stateless3/exp \
    --max-duration 600 \
    --decoding-method modified_beam_search \
    --max-sym-per-frame 1

# beam search
./pruned_transducer_stateless3/decode.py \
    --epoch 25 \
    --avg 7 \
    --exp-dir ./pruned_transducer_stateless3/exp \
    --max-duration 600 \
    --decoding-method beam_search \
    --max-sym-per-frame 1

# fast beam search
for epoch in 27; do
  for avg in 10 12; do
    ./pruned_transducer_stateless3/decode.py \
        --epoch $epoch \
        --avg $avg \
        --exp-dir ./pruned_transducer_stateless3/exp \
        --max-duration 600 \
        --decoding-method fast_beam_search \
        --max-states 32 \
        --beam 8
  done
done

The following table shows the Nbest oracle WER for fast beam search.

epoch	avg	num_paths	nbest_scale	test-clean	test-other
27	10	50	0.5	0.91	2.74
27	10	50	0.8	0.94	2.82
27	10	50	1.0	1.06	2.88
27	10	100	0.5	0.82	2.58
27	10	100	0.8	0.92	2.65
27	10	100	1.0	0.95	2.77
27	10	200	0.5	0.81	2.50
27	10	200	0.8	0.85	2.56
27	10	200	1.0	0.91	2.64
27	10	400	0.5	N/A	N/A
27	10	400	0.8	0.81	2.49
27	10	400	1.0	0.85	2.54

The Nbest oracle WER is computed using the following steps:

1. Use fast_beam_search to produce a lattice.
1. Extract N paths from the lattice using k2.random_path
1. Unique paths so that each path has a distinct sequence of tokens
1. Compute the edit distance of each path with the ground truth
1. The path with the lowest edit distance is the final output and is used to compute the WER

The command to compute the Nbest oracle WER is:

for epoch in 27; do
  for avg in 10 ; do
    for num_paths in 50 100 200 400; do
      for nbest_scale in 0.5 0.8 1.0; do
        ./pruned_transducer_stateless3/decode.py \
            --epoch $epoch \
            --avg $avg \
            --exp-dir ./pruned_transducer_stateless3/exp \
            --max-duration 600 \
            --decoding-method fast_beam_search_nbest_oracle \
            --num-paths $num_paths \
            --max-states 32 \
            --beam 8 \
            --nbest-scale $nbest_scale
      done
    done
  done
done

LibriSpeech BPE training results (Pruned Transducer 3, 2022-05-13)

Same setup as pruned_transducer_stateless3 (2022-04-29) but change --giga-prob from 0.8 to 0.9. Also use repeat on gigaspeech XL subset so that the gigaspeech dataloader never exhausts.

	test-clean	test-other	comment
greedy search (max sym per frame 1)	2.03	4.70	--iter 1224000 --avg 14 --max-duration 600
modified beam search	2.00	4.63	--iter 1224000 --avg 14 --max-duration 600
modified beam search + rnnlm shallow fusion	1.94	4.2	--iter 1224000 --avg 14 --max-duration 600
modified beam search + rnnlm + LODR	1.77	3.99	--iter 1224000 --avg 14 --max-duration 600
modified beam search + TransformerLM + LODR	1.75	3.94	--iter 1224000 --avg 14 --max-duration 600
fast beam search	2.10	4.68	--iter 1224000 --avg 14 --max-duration 600

The training commands are:

export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"

./prepare.sh
./prepare_giga_speech.sh

./pruned_transducer_stateless3/train.py \
  --world-size 8 \
  --num-epochs 30 \
  --start-epoch 0 \
  --full-libri 1 \
  --exp-dir pruned_transducer_stateless3/exp-0.9 \
  --max-duration 300 \
  --use-fp16 1 \
  --lr-epochs 4 \
  --num-workers 2 \
  --giga-prob 0.9

The tensorboard log is available at https://tensorboard.dev/experiment/HpocR7dKS9KCQkJeYxfXug/

Decoding commands:

for iter in 1224000; do
  for avg in 14; do
    for method in greedy_search modified_beam_search fast_beam_search ; do
      ./pruned_transducer_stateless3/decode.py \
        --iter $iter \
        --avg $avg \
        --exp-dir ./pruned_transducer_stateless3/exp-0.9/ \
        --max-duration 600 \
        --decoding-method $method \
        --max-sym-per-frame 1 \
        --beam 4 \
        --max-contexts 32
    done
  done
done

You may also decode using shallow fusion with external neural network LM. To do so you need to download a well-trained NN LM: RNN LM: https://huggingface.co/ezerhouni/icefall-librispeech-rnn-lm/tree/main Transformer LM: https://huggingface.co/marcoyang/icefall-librispeech-transformer-lm/tree/main

rnn_lm_scale=0.3

for iter in 1224000; do
  for avg in 14; do
    for method in modified_beam_search_rnnlm_shallow_fusion ; do
      ./pruned_transducer_stateless3/decode.py \
        --iter $iter \
        --avg $avg \
        --exp-dir ./pruned_transducer_stateless3/exp-0.9/ \
        --max-duration 600 \
        --decoding-method $method \
        --max-sym-per-frame 1 \
        --beam 4 \
        --max-contexts 32 \
        --rnn-lm-scale $rnn_lm_scale \
        --rnn-lm-exp-dir /path/to/RNNLM \
        --rnn-lm-epoch 99 \
        --rnn-lm-avg 1 \
        --rnn-lm-num-layers 3 \
        --rnn-lm-tie-weights 1
    done
  done
done

If you want to try out with LODR decoding, use the following command. This assums you have a bi-gram LM trained on LibriSpeech text. You can also download the bi-gram LM from here https://huggingface.co/marcoyang/librispeech_bigram/tree/main and put it under the directory data/lang_bpe_500.

rnn_lm_scale=0.4

for iter in 1224000; do
  for avg in 14; do
    for method in modified_beam_search_rnnlm_LODR ; do
      ./pruned_transducer_stateless3/decode.py \
        --iter $iter \
        --avg $avg \
        --exp-dir ./pruned_transducer_stateless3/exp-0.9/ \
        --max-duration 600 \
        --decoding-method $method \
        --max-sym-per-frame 1 \
        --beam 4 \
        --max-contexts 32 \
        --rnn-lm-scale $rnn_lm_scale \
        --rnn-lm-exp-dir /path/to/RNNLM \
        --rnn-lm-epoch 99 \
        --rnn-lm-avg 1 \
        --rnn-lm-num-layers 3 \
        --rnn-lm-tie-weights 1 \
        --tokens-ngram 2 \
        --ngram-lm-scale -0.14
    done
  done
done

The pretrained models, training logs, decoding logs, and decoding results can be found at https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13

LibriSpeech BPE training results (Pruned Transducer 2)

pruned_transducer_stateless2

This is with a reworked version of the conformer encoder, with many changes.

Training on full librispeech

Using commit 34aad74a2c849542dd5f6359c9e6b527e8782fd6. See #288

The WERs are:

	test-clean	test-other	comment
greedy search (max sym per frame 1)	2.62	6.37	--epoch 25 --avg 8 --max-duration 600
fast beam search	2.61	6.17	--epoch 25 --avg 8 --max-duration 600 --decoding-method fast_beam_search
modified beam search	2.59	6.19	--epoch 25 --avg 8 --max-duration 600 --decoding-method modified_beam_search
greedy search (max sym per frame 1)	2.70	6.04	--epoch 34 --avg 10 --max-duration 600
fast beam search	2.66	6.00	--epoch 34 --avg 10 --max-duration 600 --decoding-method fast_beam_search
greedy search (max sym per frame 1)	2.62	6.03	--epoch 38 --avg 10 --max-duration 600
fast beam search	2.57	5.95	--epoch 38 --avg 10 --max-duration 600 --decoding-method fast_beam_search

The train and decode commands are:

python3 ./pruned_transducer_stateless2/train.py \
  --exp-dir=pruned_transducer_stateless2/exp \
  --world-size 8 \
  --num-epochs 26 \
  --full-libri 1 \
  --max-duration 300

and:

python3 ./pruned_transducer_stateless2/decode.py \
  --exp-dir pruned_transducer_stateless2/exp \
  --epoch 25 \
  --avg 8 \
  --bpe-model ./data/lang_bpe_500/bpe.model \
  --max-duration 600

The Tensorboard log is at https://tensorboard.dev/experiment/Xoz0oABMTWewo1slNFXkyA (apologies, log starts only from epoch 3).

The pretrained models, training logs, decoding logs, and decoding results can be found at https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless2-2022-04-29

Training on train-clean-100:

Trained with 1 job:

python3 ./pruned_transducer_stateless2/train.py \
  --exp-dir=pruned_transducer_stateless2/exp_100h_ws1 \
  --world-size 1 \
  --num-epochs 40  \
  --full-libri 0 \
  --max-duration 300

and decoded with:

python3 ./pruned_transducer_stateless2/decode.py \
  --exp-dir pruned_transducer_stateless2/exp_100h_ws1 \
  --epoch 19 \
  --avg 8 \
  --bpe-model ./data/lang_bpe_500/bpe.model \
  --max-duration 600

The Tensorboard log is at https://tensorboard.dev/experiment/AhnhooUBRPqTnaggoqo7lg (learning rate schedule is not visible due to a since-fixed bug).

	test-clean	test-other	comment
greedy search (max sym per frame 1)	7.12	18.42	--epoch 19 --avg 8
greedy search (max sym per frame 1)	6.71	17.77	--epoch 29 --avg 8
greedy search (max sym per frame 1)	6.64	17.19	--epoch 39 --avg 10
fast beam search	6.58	17.27	--epoch 29 --avg 8 --decoding-method fast_beam_search
fast beam search	6.53	16.82	--epoch 39 --avg 10 --decoding-method fast_beam_search

Trained with 2 jobs:

python3 ./pruned_transducer_stateless2/train.py \
  --exp-dir=pruned_transducer_stateless2/exp_100h_ws2 \
  --world-size 2 \
  --num-epochs 40  \
  --full-libri 0 \
  --max-duration 300

and decoded with:

python3 ./pruned_transducer_stateless2/decode.py \
  --exp-dir pruned_transducer_stateless2/exp_100h_ws2 \
  --epoch 19 \
  --avg 8 \
  --bpe-model ./data/lang_bpe_500/bpe.model \
  --max-duration 600

The Tensorboard log is at https://tensorboard.dev/experiment/dvOC9wsrSdWrAIdsebJILg/ (learning rate schedule is not visible due to a since-fixed bug).

	test-clean	test-other	comment
greedy search (max sym per frame 1)	7.05	18.77	--epoch 19 --avg 8
greedy search (max sym per frame 1)	6.82	18.14	--epoch 29 --avg 8
greedy search (max sym per frame 1)	6.81	17.66	--epoch 30 --avg 10

Trained with 4 jobs:

python3 ./pruned_transducer_stateless2/train.py \
  --exp-dir=pruned_transducer_stateless2/exp_100h_ws4 \
  --world-size 4 \
  --num-epochs 40  \
  --full-libri 0 \
  --max-duration 300

and decoded with:

python3 ./pruned_transducer_stateless2/decode.py \
  --exp-dir pruned_transducer_stateless2/exp_100h_ws4 \
  --epoch 19 \
  --avg 8 \
  --bpe-model ./data/lang_bpe_500/bpe.model \
  --max-duration 600

The Tensorboard log is at https://tensorboard.dev/experiment/a3T0TyC0R5aLj5bmFbRErA/ (learning rate schedule is not visible due to a since-fixed bug).

	test-clean	test-other	comment
greedy search (max sym per frame 1)	7.31	19.55	--epoch 19 --avg 8
greedy search (max sym per frame 1)	7.08	18.59	--epoch 29 --avg 8
greedy search (max sym per frame 1)	6.86	18.29	--epoch 30 --avg 10

Trained with 1 job, with --use-fp16=True --max-duration=300 i.e. with half-precision floats (but without increasing max-duration), after merging #305. Train command was

python3 ./pruned_transducer_stateless2/train.py \
  --exp-dir=pruned_transducer_stateless2/exp_100h_fp16 \
  --world-size 1 \
  --num-epochs 40  \
  --full-libri 0 \
  --max-duration 300 \
  --use-fp16 True

The Tensorboard log is at https://tensorboard.dev/experiment/DAtGG9lpQJCROUDwPNxwpA

	test-clean	test-other	comment
greedy search (max sym per frame 1)	7.10	18.57	--epoch 19 --avg 8
greedy search (max sym per frame 1)	6.81	17.84	--epoch 29 --avg 8
greedy search (max sym per frame 1)	6.63	17.39	--epoch 30 --avg 10

Trained with 1 job, with --use-fp16=True --max-duration=500, i.e. with half-precision floats and max-duration increased from 300 to 500, after merging #305. Train command was

python3 ./pruned_transducer_stateless2/train.py \
  --exp-dir=pruned_transducer_stateless2/exp_100h_fp16 \
  --world-size 1 \
  --num-epochs 40  \
  --full-libri 0 \
  --max-duration 500 \
  --use-fp16 True

The Tensorboard log is at https://tensorboard.dev/experiment/Km7QBHYnSLWs4qQnAJWsaA

	test-clean	test-other	comment
greedy search (max sym per frame 1)	7.10	18.79	--epoch 19 --avg 8
greedy search (max sym per frame 1)	6.92	18.16	--epoch 29 --avg 8
greedy search (max sym per frame 1)	6.89	17.75	--epoch 30 --avg 10

LibriSpeech BPE training results (Pruned Transducer)

Conformer encoder + non-current decoder. The decoder contains only an embedding layer, a Conv1d (with kernel size 2) and a linear layer (to transform tensor dim).

2022-03-12

pruned_transducer_stateless

Using commit 1603744469d167d848e074f2ea98c587153205fa. See #248

The WERs are:

	test-clean	test-other	comment
greedy search (max sym per frame 1)	2.62	6.37	--epoch 42 --avg 11 --max-duration 100
greedy search (max sym per frame 2)	2.62	6.37	--epoch 42 --avg 11 --max-duration 100
greedy search (max sym per frame 3)	2.62	6.37	--epoch 42 --avg 11 --max-duration 100
modified beam search (beam size 4)	2.56	6.27	--epoch 42 --avg 11 --max-duration 100
beam search (beam size 4)	2.57	6.27	--epoch 42 --avg 11 --max-duration 100

The decoding time for test-clean and test-other is given below: (A V100 GPU with 32 GB RAM is used for decoding. Note: Not all GPU RAM is used during decoding.)

decoding method	test-clean (seconds)	test-other (seconds)
greedy search (--max-sym-per-frame=1)	160	159
greedy search (--max-sym-per-frame=2)	184	177
greedy search (--max-sym-per-frame=3)	210	213
modified beam search (--beam-size 4)	273	269
beam search (--beam-size 4)	2741	2221

We recommend you to use modified_beam_search.

Training command:

cd egs/librispeech/ASR/
./prepare.sh

export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"

. path.sh

./pruned_transducer_stateless/train.py \
  --world-size 8 \
  --num-epochs 60 \
  --start-epoch 0 \
  --exp-dir pruned_transducer_stateless/exp \
  --full-libri 1 \
  --max-duration 300 \
  --prune-range 5 \
  --lr-factor 5 \
  --lm-scale 0.25

The tensorboard training log can be found at https://tensorboard.dev/experiment/WKRFY5fYSzaVBHahenpNlA/

The command for decoding is:

epoch=42
avg=11
sym=1

# greedy search

./pruned_transducer_stateless/decode.py \
  --epoch $epoch \
  --avg $avg \
  --exp-dir ./pruned_transducer_stateless/exp \
  --max-duration 100 \
  --decoding-method greedy_search \
  --beam-size 4 \
  --max-sym-per-frame $sym

# modified beam search
./pruned_transducer_stateless/decode.py \
  --epoch $epoch \
  --avg $avg \
  --exp-dir ./pruned_transducer_stateless/exp \
  --max-duration 100 \
  --decoding-method modified_beam_search \
  --beam-size 4

# beam search
# (not recommended)
./pruned_transducer_stateless/decode.py \
  --epoch $epoch \
  --avg $avg \
  --exp-dir ./pruned_transducer_stateless/exp \
  --max-duration 100 \
  --decoding-method beam_search \
  --beam-size 4

You can find a pre-trained model, decoding logs, and decoding results at https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless-2022-03-12

2022-02-18

pruned_transducer_stateless

The WERs are

	test-clean	test-other	comment
greedy search	2.85	6.98	--epoch 28 --avg 15 --max-duration 100

The training command for reproducing is given below:

export CUDA_VISIBLE_DEVICES="0,1,2,3"

./pruned_transducer_stateless/train.py \
  --world-size 4 \
  --num-epochs 30 \
  --start-epoch 0 \
  --exp-dir pruned_transducer_stateless/exp \
  --full-libri 1 \
  --max-duration 300 \
  --prune-range 5 \
  --lr-factor 5 \
  --lm-scale 0.25 \

The tensorboard training log can be found at https://tensorboard.dev/experiment/ejG7VpakRYePNNj6AbDEUw/#scalars

The decoding command is:

epoch=28
avg=15

## greedy search
./pruned_transducer_stateless/decode.py \
  --epoch $epoch \
  --avg $avg \
  --exp-dir pruned_transducer_stateless/exp \
  --max-duration 100

LibriSpeech BPE training results (Transducer)

Conformer encoder + embedding decoder

Conformer encoder + non-recurrent decoder. The decoder contains only an embedding layer and a Conv1d (with kernel size 2).

See

./transducer_stateless
./transducer_stateless_multi_datasets

2022-03-01

Using commit 2332ba312d7ce72f08c7bac1e3312f7e3dd722dc.

It uses GigaSpeech as extra training data. 20% of the time it selects a batch from L subset of GigaSpeech and 80% of the time it selects a batch from LibriSpeech.

The WERs are

	test-clean	test-other	comment
greedy search (max sym per frame 1)	2.64	6.55	--epoch 39 --avg 15 --max-duration 100
modified beam search (beam size 4)	2.61	6.46	--epoch 39 --avg 15 --max-duration 100

The training command for reproducing is given below:

cd egs/librispeech/ASR/
./prepare.sh
./prepare_giga_speech.sh

export CUDA_VISIBLE_DEVICES="0,1,2,3"

./transducer_stateless_multi_datasets/train.py \
  --world-size 4 \
  --num-epochs 40 \
  --start-epoch 0 \
  --exp-dir transducer_stateless_multi_datasets/exp-full-2 \
  --full-libri 1 \
  --max-duration 300 \
  --lr-factor 5 \
  --bpe-model data/lang_bpe_500/bpe.model \
  --modified-transducer-prob 0.25 \
  --giga-prob 0.2

The tensorboard training log can be found at https://tensorboard.dev/experiment/xmo5oCgrRVelH9dCeOkYBg/

The decoding command is:

epoch=39
avg=15
sym=1

# greedy search
./transducer_stateless_multi_datasets/decode.py \
  --epoch $epoch \
  --avg $avg \
  --exp-dir transducer_stateless_multi_datasets/exp-full-2 \
  --bpe-model ./data/lang_bpe_500/bpe.model \
  --max-duration 100 \
  --context-size 2 \
  --max-sym-per-frame $sym

# modified beam search
./transducer_stateless_multi_datasets/decode.py \
  --epoch $epoch \
  --avg $avg \
  --exp-dir transducer_stateless_multi_datasets/exp-full-2 \
  --bpe-model ./data/lang_bpe_500/bpe.model \
  --max-duration 100 \
  --context-size 2 \
  --decoding-method modified_beam_search \
  --beam-size 4

You can find a pretrained model by visiting https://huggingface.co/csukuangfj/icefall-asr-librispeech-transducer-stateless-multi-datasets-bpe-500-2022-03-01

2022-04-19

transducer_stateless2

This version uses torchaudio's RNN-T loss.

Using commit fce7f3cd9a486405ee008bcbe4999264f27774a3. See #316

	test-clean	test-other	comment
greedy search (max sym per frame 1)	2.65	6.30	--epoch 59 --avg 10 --max-duration 600
greedy search (max sym per frame 2)	2.62	6.23	--epoch 59 --avg 10 --max-duration 100
greedy search (max sym per frame 3)	2.62	6.23	--epoch 59 --avg 10 --max-duration 100
modified beam search	2.63	6.15	--epoch 59 --avg 10 --max-duration 100 --decoding-method modified_beam_search
beam search	2.59	6.15	--epoch 59 --avg 10 --max-duration 100 --decoding-method beam_search

Note: This model is trained with standard RNN-T loss. Neither modified transducer nor pruned RNN-T is used. You can see that there is a performance degradation in WER when we limit the max symbol per frame to 1.

The number of active paths in modified_beam_search and beam_search is 4.

The training and decoding commands are:

export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"

./transducer_stateless2/train.py \
  --world-size 8 \
  --num-epochs 60 \
  --start-epoch 0 \
  --exp-dir transducer_stateless2/exp-2 \
  --full-libri 1 \
  --max-duration 300 \
  --lr-factor 5

epoch=59
avg=10
# greedy search
./transducer_stateless2/decode.py \
  --epoch $epoch \
  --avg $avg \
  --exp-dir ./transducer_stateless2/exp-2 \
  --max-duration 600 \
  --decoding-method greedy_search \
  --max-sym-per-frame 1

# modified beam search
./transducer_stateless2/decode.py \
  --epoch $epoch \
  --avg $avg \
  --exp-dir ./transducer_stateless2/exp-2 \
  --max-duration 100 \
  --decoding-method modified_beam_search \

# beam search
./transducer_stateless2/decode.py \
  --epoch $epoch \
  --avg $avg \
  --exp-dir ./transducer_stateless2/exp-2 \
  --max-duration 100 \
  --decoding-method beam_search \

The tensorboard log is at https://tensorboard.dev/experiment/oAlle3dxQD2EY8ePwjIGuw/.

You can find a pre-trained model, decoding logs, and decoding results at https://huggingface.co/csukuangfj/icefall-asr-librispeech-transducer-stateless2-torchaudio-2022-04-19

2022-02-07

Using commit a8150021e01d34ecbd6198fe03a57eacf47a16f2.

The WERs are

	test-clean	test-other	comment
greedy search (max sym per frame 1)	2.67	6.67	--epoch 63 --avg 19 --max-duration 100
greedy search (max sym per frame 2)	2.67	6.67	--epoch 63 --avg 19 --max-duration 100
greedy search (max sym per frame 3)	2.67	6.67	--epoch 63 --avg 19 --max-duration 100
modified beam search (beam size 4)	2.67	6.57	--epoch 63 --avg 19 --max-duration 100

The training command for reproducing is given below:

cd egs/librispeech/ASR/
./prepare.sh
export CUDA_VISIBLE_DEVICES="0,1,2,3"
./transducer_stateless/train.py \
  --world-size 4 \
  --num-epochs 76 \
  --start-epoch 0 \
  --exp-dir transducer_stateless/exp-full \
  --full-libri 1 \
  --max-duration 300 \
  --lr-factor 5 \
  --bpe-model data/lang_bpe_500/bpe.model \
  --modified-transducer-prob 0.25

The tensorboard training log can be found at https://tensorboard.dev/experiment/qgvWkbF2R46FYA6ZMNmOjA/#scalars

The decoding command is:

epoch=63
avg=19

## greedy search
for sym in 1 2 3; do
  ./transducer_stateless/decode.py \
    --epoch $epoch \
    --avg $avg \
    --exp-dir transducer_stateless/exp-full \
    --bpe-model ./data/lang_bpe_500/bpe.model \
    --max-duration 100 \
    --max-sym-per-frame $sym
done

## modified beam search

./transducer_stateless/decode.py \
  --epoch $epoch \
  --avg $avg \
  --exp-dir transducer_stateless/exp-full \
  --bpe-model ./data/lang_bpe_500/bpe.model \
  --max-duration 100 \
  --context-size 2 \
  --decoding-method modified_beam_search \
  --beam-size 4

You can find a pretrained model by visiting https://huggingface.co/csukuangfj/icefall-asr-librispeech-transducer-stateless-bpe-500-2022-02-07

Conformer encoder + LSTM decoder

Using commit 8187d6236c2926500da5ee854f758e621df803cc.

Conformer encoder + LSTM decoder.

The best WER is

	test-clean	test-other
WER	3.07	7.51

using --epoch 34 --avg 11 with greedy search.

The training command to reproduce the above WER is:

export CUDA_VISIBLE_DEVICES="0,1,2,3"

./transducer/train.py \
  --world-size 4 \
  --num-epochs 35 \
  --start-epoch 0 \
  --exp-dir transducer/exp-lr-2.5-full \
  --full-libri 1 \
  --max-duration 180 \
  --lr-factor 2.5

The decoding command is:

epoch=34
avg=11

./transducer/decode.py \
  --epoch $epoch \
  --avg $avg \
  --exp-dir transducer/exp-lr-2.5-full \
  --bpe-model ./data/lang_bpe_500/bpe.model \
  --max-duration 100

You can find the tensorboard log at: https://tensorboard.dev/experiment/D7NQc3xqTpyVmWi5FnWjrA

LibriSpeech BPE training results (Conformer-CTC 2)

conformer_ctc2

2022-07-21

It implements a 'reworked' version of CTC attention model. As demenstrated by pruned_transducer_stateless2, reworked Conformer model has superior performance compared to the original Conformer. So in this modified version of CTC attention model, it has the reworked Conformer as the encoder and the reworked Transformer as the decoder. conformer_ctc2 also integrates with the idea of the 'averaging models' in pruned_transducer_stateless4.

The WERs on comparisons with a baseline model, for the librispeech test dataset, are listed as below.

The baseline model is the original conformer CTC attention model trained with icefall/egs/librispeech/ASR/conformer_ctc. The model is downloaded from https://huggingface.co/csukuangfj/icefall-asr-librispeech-conformer-ctc-jit-bpe-500-2021-11-09. This model has 12 layers of Conformer encoder layers and 6 Transformer decoder layers. Number of model parameters is 109,226,120. It has been trained with 90 epochs with full Librispeech dataset.

For this reworked CTC attention model, it has 12 layers of reworked Conformer encoder layers and 6 reworked Transformer decoder layers. Number of model parameters is 103,071,035. With full Librispeech data set, it was trained for only 30 epochs because the reworked model would converge much faster. Please refer to https://tensorboard.dev/experiment/GR1s6VrJRTW5rtB50jakew/#scalars to see the loss convergence curve. Please find the above trained model at https://huggingface.co/WayneWiser/icefall-asr-librispeech-conformer-ctc2-jit-bpe-500-2022-07-21 in huggingface.

The decoding configuration for the reworked model is --epoch 30, --avg 8, --use-averaged-model True, which is the best after searching.

WER	reworked ctc attention	with --epoch 30 --avg 8 --use-averaged-model True		ctc attention	with --epoch 77 --avg 55
test sets	test-clean	test-other	Avg	test-clean	test-other	Avg
ctc-greedy-search	2.98%	7.14%	5.06%	2.90%	7.47%	5.19%
ctc-decoding	2.98%	7.14%	5.06%	2.90%	7.47%	5.19%
1best	2.93%	6.37%	4.65%	2.70%	6.49%	4.60%
nbest	2.94%	6.39%	4.67%	2.70%	6.48%	4.59%
nbest-rescoring	2.68%	5.77%	4.23%	2.55%	6.07%	4.31%
whole-lattice-rescoring	2.66%	5.76%	4.21%	2.56%	6.04%	4.30%
attention-decoder	2.59%	5.54%	4.07%	2.41%	5.77%	4.09%
nbest-oracle	1.53%	3.47%	2.50%	1.69%	4.02%	2.86%
rnn-lm	2.37%	4.98%	3.68%	2.31%	5.35%	3.83%

conformer_ctc2 also implements the CTC greedy search decoding, it has the identical WERs with the CTC-decoding method. For other decoding methods, the average WER of the two test sets with the two models is similar. Except for the 1best and nbest methods, the overall performance of reworked model is better than the baseline model.

To reproduce the above result, use the following commands.

The training commands are:

    WORLD_SIZE=8
    export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5,6,7"
    ./conformer_ctc2/train.py \
    --manifest-dir data/fbank \
    --exp-dir conformer_ctc2/exp \
    --full-libri 1 \
    --spec-aug-time-warp-factor 80 \
    --max-duration 300 \
    --world-size ${WORLD_SIZE} \
    --start-epoch 1 \
    --num-epochs 30 \
    --att-rate 0.7 \
    --num-decoder-layers 6

And the following commands are for decoding:

for method in ctc-greedy-search ctc-decoding 1best nbest-oracle; do
  python3 ./conformer_ctc2/decode.py \
  --exp-dir conformer_ctc2/exp \
  --use-averaged-model True --epoch 30 --avg 8 --max-duration 200 --method $method
done

for method in nbest nbest-rescoring whole-lattice-rescoring attention-decoder ; do
  python3 ./conformer_ctc2/decode.py \
  --exp-dir conformer_ctc2/exp \
  --use-averaged-model True --epoch 30 --avg 8 --max-duration 20 --method $method
done

rnn_dir=$(git rev-parse --show-toplevel)/icefall/rnn_lm
./conformer_ctc2/decode.py \
  --exp-dir conformer_ctc2/exp \
  --lang-dir data/lang_bpe_500 \
  --lm-dir data/lm \
  --max-duration 30 \
  --concatenate-cuts 0 \
  --bucketing-sampler 1 \
  --num-paths 1000 \
  --use-averaged-model True \
  --epoch 30 \
  --avg 8 \
  --nbest-scale 0.5 \
  --rnn-lm-exp-dir ${rnn_dir}/exp \
  --rnn-lm-epoch 29 \
  --rnn-lm-avg 3 \
  --rnn-lm-embedding-dim 2048 \
  --rnn-lm-hidden-dim 2048 \
  --rnn-lm-num-layers 3 \
  --rnn-lm-tie-weights true \
  --method rnn-lm

You can find the RNN-LM pre-trained model at https://huggingface.co/ezerhouni/icefall-librispeech-rnn-lm/tree/main

LibriSpeech BPE training results (Conformer-CTC)

2021-11-09

The best WER, as of 2022-06-20, for the librispeech test dataset is below (using HLG decoding + n-gram LM rescoring + attention decoder rescoring + rnn lm rescoring):

	test-clean	test-other
WER	2.32	5.39

Scale values used in n-gram LM rescoring and attention rescoring for the best WERs are:

ngram_lm_scale	attention_scale	rnn_lm_scale
0.3	2.1	2.2

To reproduce the above result, use the following commands for training:

cd egs/librispeech/ASR/conformer_ctc
./prepare.sh
export CUDA_VISIBLE_DEVICES="0,1,2,3"
./conformer_ctc/train.py \
  --exp-dir conformer_ctc/exp_500_att0.8 \
  --lang-dir data/lang_bpe_500 \
  --att-rate 0.8 \
  --full-libri 1 \
  --max-duration 200 \
  --concatenate-cuts 0 \
  --world-size 4 \
  --bucketing-sampler 1 \
  --start-epoch 0 \
  --num-epochs 90
# Note: It trains for 90 epochs, but the best WER is at epoch-77.pt

# Train the RNN-LM
cd icefall
export CUDA_VISIBLE_DEVICES="0,1,2,3"
./rnn_lm/train.py \
  --exp-dir rnn_lm/exp_2048_3_tied \
  --start-epoch 0 \
  --world-size 4 \
  --num-epochs 30 \
  --use-fp16 1 \
  --embedding-dim 2048 \
  --hidden-dim 2048 \
  --num-layers 3 \
  --batch-size 500 \
  --tie-weights true

and the following command for decoding

rnn_dir=$(git rev-parse --show-toplevel)/icefall/rnn_lm
./conformer_ctc/decode.py \
  --exp-dir conformer_ctc/exp_500_att0.8 \
  --lang-dir data/lang_bpe_500 \
  --max-duration 30 \
  --concatenate-cuts 0 \
  --bucketing-sampler 1 \
  --num-paths 1000 \
  --epoch 77 \
  --avg 55 \
  --nbest-scale 0.5 \
  --rnn-lm-exp-dir ${rnn_dir}/exp_2048_3_tied \
  --rnn-lm-epoch 29 \
  --rnn-lm-avg 3 \
  --rnn-lm-embedding-dim 2048 \
  --rnn-lm-hidden-dim 2048 \
  --rnn-lm-num-layers 3 \
  --rnn-lm-tie-weights true \
  --method rnn-lm

You can find the Conformer-CTC pre-trained model by visiting https://huggingface.co/csukuangfj/icefall-asr-librispeech-conformer-ctc-jit-bpe-500-2021-11-09

and the RNN-LM pre-trained model: https://huggingface.co/ezerhouni/icefall-librispeech-rnn-lm/tree/main

The tensorboard log for training is available at https://tensorboard.dev/experiment/hZDWrZfaSqOMqtW0NEfXKg/#scalars

2021-08-19

(Wei Kang): Result of #13

TensorBoard log is available at https://tensorboard.dev/experiment/GnRzq8WWQW62dK4bklXBTg/#scalars

Pretrained model is available at https://huggingface.co/pkufool/icefall_asr_librispeech_conformer_ctc

The best decoding results (WER) are listed below, we got this results by averaging models from epoch 15 to 34, and using attention-decoder decoder with num_paths equals to 100.

	test-clean	test-other
WER	2.57%	5.94%

To get more unique paths, we scaled the lattice.scores with 0.5 (see #10 (comment) for more details), we searched the lm_score_scale and attention_score_scale for best results, the scales that produced the WER above are also listed below.

	lm_scale	attention_scale
test-clean	1.3	1.2
test-other	1.2	1.1

You can use the following commands to reproduce our results:

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

# It was using ef233486, you may not need to switch to it
# git checkout ef233486

cd egs/librispeech/ASR
./prepare.sh

export CUDA_VISIBLE_DEVICES="0,1,2,3"
python conformer_ctc/train.py --bucketing-sampler True \
                              --concatenate-cuts False \
                              --max-duration 200 \
                              --full-libri True \
                              --world-size 4 \
                              --lang-dir data/lang_bpe_5000

python conformer_ctc/decode.py --nbest-scale 0.5 \
                               --epoch 34 \
                               --avg 20 \
                               --method attention-decoder \
                               --max-duration 20 \
                               --num-paths 100 \
                               --lang-dir data/lang_bpe_5000

LibriSpeech training results (Tdnn-Lstm)

2021-08-24

(Wei Kang): Result of phone based Tdnn-Lstm model.

Icefall version: https://github.com/k2-fsa/icefall/commit/caa0b9e9425af27e0c6211048acb55a76ed5d315

Pretrained model is available at https://huggingface.co/pkufool/icefall_asr_librispeech_tdnn-lstm_ctc

The best decoding results (WER) are listed below, we got this results by averaging models from epoch 19 to 14, and using whole-lattice-rescoring decoding method.

	test-clean	test-other
WER	6.59%	17.69%

We searched the lm_score_scale for best results, the scales that produced the WER above are also listed below.

	lm_scale
test-clean	0.8
test-other	0.9

Files

RESULTS.md

Latest commit

History

RESULTS.md

File metadata and controls

Results

zipformer (zipformer + pruned stateless transducer + CTC)

Non-streaming

normal-scaled model, number of model parameters: 65805511, i.e., 65.81 M

zipformer (zipformer + pruned stateless transducer)

Non-streaming

normal-scaled model, number of model parameters: 65549011, i.e., 65.55 M

small-scaled model, number of model parameters: 23285615, i.e., 23.3 M

large-scaled model, number of model parameters: 148439574, i.e., 148.4 M

large-scaled model, number of model parameters: 148439574, i.e., 148.4 M, trained on 8 80G-A100 GPUs

streaming

normal-scaled model, number of model parameters: 66110931, i.e., 66.11 M

Zipformer CTC

pruned_transducer_stateless7 (Fine-tune with mux)

pruned_transducer_stateless7 (zipformer + multidataset(LibriSpeech + GigaSpeech + CommonVoice 13.0))

Streaming Zipformer-Transducer (Pruned Stateless Transducer + Streaming Zipformer + Multi-Dataset)

training on full librispeech + full gigaspeech (with giga_prob=0.9)

Smaller model

Streaming Zipformer-Transducer (Pruned Stateless Transducer + Streaming Zipformer)

training on full librispeech

Smaller model

zipformer_mmi (zipformer with mmi loss)

pruned_transducer_stateless7_ctc_bs (zipformer with transducer loss and ctc loss using blank skip)

greedy_search

modified_beam_search

pruned_transducer_stateless7_ctc (zipformer with transducer loss and ctc loss)

LibriSpeech BPE training results (Conformer CTC, supporting delay penalty)

training on full librispeech

pruned_transducer_stateless8 (zipformer + multidataset)

pruned_transducer_stateless7 (zipformer)

LibriSpeech BPE training results (Pruned Stateless LSTM RNN-T + gradient filter)

training on full librispeech

LibriSpeech BPE training results (Pruned Stateless LSTM RNN-T + multi-dataset)

LibriSpeech BPE training results (Pruned Stateless LSTM RNN-T)

training on full librispeech

LibriSpeech BPE training results (Pruned Stateless Conv-Emformer RNN-T 2)

With lower latency setup, training on full librispeech

With higher latency setup, training on full librispeech

LibriSpeech BPE training results (Pruned Stateless Streaming Conformer RNN-T)

Training on full librispeech

Training on full librispeech

Training on full librispeech (Use giga_prob = 0.5)

Training on full librispeech (Use giga_prob = 0.9)

Training on full librispeech

Training on full librispeech

LibriSpeech BPE training results (Pruned Stateless Conv-Emformer RNN-T)

Training on full librispeech

LibriSpeech BPE training results (Pruned Stateless Emformer RNN-T)

LibriSpeech BPE training results (Pruned Stateless Transducer 5)

Large

Medium

Baseline-2

LibriSpeech BPE training results (Pruned Stateless Transducer 4)

Training on full librispeech

Training on train-clean-100

LibriSpeech BPE training results (Pruned Stateless Transducer 3, 2022-04-29)

LibriSpeech BPE training results (Pruned Transducer 3, 2022-05-13)

LibriSpeech BPE training results (Pruned Transducer 2)

Training on full librispeech

Training on train-clean-100:

LibriSpeech BPE training results (Pruned Transducer)

2022-03-12

2022-02-18

LibriSpeech BPE training results (Transducer)

Conformer encoder + embedding decoder

2022-03-01

2022-04-19

2022-02-07

Conformer encoder + LSTM decoder

LibriSpeech BPE training results (Conformer-CTC 2)

2022-07-21

LibriSpeech BPE training results (Conformer-CTC)

2021-11-09

2021-08-19