export CUDA_VISIBLE_DEVICES="0,1"
# for non-streaming model training:
./zipformer/train.py \
  --world-size 2 \
  --num-epochs 50 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir zipformer/exp-large-cr-ctc-rnnt \
  --use-cr-ctc 1 \
  --use-ctc 1 \
  --use-transducer 1 \
  --use-attention-decoder 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 \
  --ctc-loss-scale 0.1 \
  --enable-spec-aug 0 \
  --cr-loss-scale 0.02 \
  --time-mask-ratio 2.5 \
  --full-libri 1 \
  --max-duration 1400 \
  --master-port 12345

export CUDA_VISIBLE_DEVICES="0"
for m in greedy_search modified_beam_search; do
  ./zipformer/decode.py \
    --epoch 50 \
    --avg 26 \
    --exp-dir zipformer/exp-large-cr-ctc-rnnt \
    --use-cr-ctc 1 \
    --use-ctc 1 \
    --use-transducer 1 \
    --use-attention-decoder 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 \
    --max-duration 300 \
    --decoding-method $m
done

export CUDA_VISIBLE_DEVICES="0,1"
# for non-streaming model training:
./zipformer/train.py \
  --world-size 2 \
  --num-epochs 50 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir zipformer/exp-large-cr-ctc-aed \
  --use-cr-ctc 1 \
  --use-ctc 1 \
  --use-transducer 0 \
  --use-attention-decoder 1 \
  --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 \
  --ctc-loss-scale 0.1 \
  --attention-decoder-loss-scale 0.9 \
  --enable-spec-aug 0 \
  --cr-loss-scale 0.02 \
  --time-mask-ratio 2.5 \
  --full-libri 1 \
  --max-duration 1200 \
  --master-port 12345

export CUDA_VISIBLE_DEVICES="0"
./zipformer/ctc_decode.py \
  --epoch 50 \
  --avg 20 \
  --exp-dir zipformer/exp-large-cr-ctc-aed/ \
  --use-cr-ctc 1 \
  --use-ctc 1 \
  --use-transducer 0 \
  --use-attention-decoder 1 \
  --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 \
  --max-duration 200 \
  --decoding-method attention-decoder-rescoring-no-ngram
done

export CUDA_VISIBLE_DEVICES="0,1"
# for non-streaming model training:
./zipformer/train.py \
  --world-size 2 \
  --num-epochs 50 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir zipformer/exp-small/ \
  --use-cr-ctc 1 \
  --use-ctc 1 \
  --use-transducer 0 \
  --use-attention-decoder 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 \
  --enable-spec-aug 0 \
  --cr-loss-scale 0.2 \
  --time-mask-ratio 2.5 \
  --full-libri 1 \
  --max-duration 850 \
  --master-port 12345

export CUDA_VISIBLE_DEVICES="0"
for m in ctc-greedy-search ctc-prefix-beam-search; do
  ./zipformer/ctc_decode.py \
    --epoch 50 \
    --avg 25 \
    --exp-dir zipformer/exp-small \
    --use-cr-ctc 1 \
    --use-ctc 1 \
    --use-transducer 0 \
    --use-attention-decoder 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 \
    --max-duration 600 \
    --decoding-method $m
done

export CUDA_VISIBLE_DEVICES="0,1,2,3"
# For non-streaming model training:
./zipformer/train.py \
  --world-size 4 \
  --num-epochs 50 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir zipformer/exp \
  --use-cr-ctc 1 \
  --use-ctc 1 \
  --use-transducer 0 \
  --use-attention-decoder 0 \
  --enable-spec-aug 0 \
  --cr-loss-scale 0.2 \
  --time-mask-ratio 2.5 \
  --full-libri 1 \
  --max-duration 700 \
  --master-port 12345

export CUDA_VISIBLE_DEVICES="0"
for m in ctc-greedy-search ctc-prefix-beam-search; do
  ./zipformer/ctc_decode.py \
    --epoch 50 \
    --avg 24 \
    --exp-dir zipformer/exp \
    --use-cr-ctc 1 \
    --use-ctc 1 \
    --use-transducer 0 \
    --use-attention-decoder 0 \
    --max-duration 600 \
    --decoding-method $m
done

export CUDA_VISIBLE_DEVICES="0,1"
# For non-streaming model training:
./zipformer/train.py \
  --world-size 2 \
  --num-epochs 50 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir zipformer/exp-large \
  --use-cr-ctc 1 \
  --use-ctc 1 \
  --use-transducer 0 \
  --use-attention-decoder 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 \
  --enable-spec-aug 0 \
  --cr-loss-scale 0.2 \
  --time-mask-ratio 2.5 \
  --full-libri 1 \
  --max-duration 1400 \
  --master-port 12345

export CUDA_VISIBLE_DEVICES="0"
for m in ctc-greedy-search ctc-prefix-beam-search; do
  ./zipformer/ctc_decode.py \
    --epoch 50 \
    --avg 26 \
    --exp-dir zipformer/exp-large \
    --use-cr-ctc 1 \
    --use-ctc 1 \
    --use-transducer 0 \
    --use-attention-decoder 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 \
    --max-duration 600 \
    --decoding-method $m
done

export CUDA_VISIBLE_DEVICES="0,1"
# For non-streaming model training:
./zipformer/train.py \
  --world-size 2 \
  --num-epochs 50 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir zipformer/exp-small \
  --full-libri 1 \
  --use-ctc 1 \
  --use-transducer 0 \
  --use-attention-decoder 1 \
  --ctc-loss-scale 0.1 \
  --attention-decoder-loss-scale 0.9 \
  --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 \
  --max-duration 1700 \
  --master-port 12345

export CUDA_VISIBLE_DEVICES="0"
for m in ctc-decoding attention-decoder-rescoring-no-ngram; do
  ./zipformer/ctc_decode.py \
    --epoch 50 \
    --avg 30 \
    --exp-dir zipformer/exp-small \
    --use-ctc 1 \
    --use-transducer 0 \
    --use-attention-decoder 1 \
    --attention-decoder-loss-scale 0.9 \
    --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 \
    --max-duration 100 \
    --causal 0 \
    --num-paths 100 \
    --decoding-method $m
done

export CUDA_VISIBLE_DEVICES="0,1,2,3"
# For non-streaming model training:
./zipformer/train.py \
  --world-size 4 \
  --num-epochs 50 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir zipformer/exp \
  --full-libri 1 \
  --use-ctc 1 \
  --use-transducer 0 \
  --use-attention-decoder 1 \
  --ctc-loss-scale 0.1 \
  --attention-decoder-loss-scale 0.9 \
  --max-duration 1200 \
  --master-port 12345

export CUDA_VISIBLE_DEVICES="0"
for m in ctc-decoding attention-decoder-rescoring-no-ngram; do
  ./zipformer/ctc_decode.py \
    --epoch 50 \
    --avg 22 \
    --exp-dir zipformer/exp \
    --use-ctc 1 \
    --use-transducer 0 \
    --use-attention-decoder 1 \
    --attention-decoder-loss-scale 0.9 \
    --max-duration 100 \
    --causal 0 \
    --num-paths 100 \
    --decoding-method $m
done

export CUDA_VISIBLE_DEVICES="0,1,2,3"
# For non-streaming model training:
./zipformer/train.py \
  --world-size 4 \
  --num-epochs 50 \
  --start-epoch 1 \
  --use-fp16 1 \
  --exp-dir zipformer/exp-large \
  --full-libri 1 \
  --use-ctc 1 \
  --use-transducer 0 \
  --use-attention-decoder 1 \
  --ctc-loss-scale 0.1 \
  --attention-decoder-loss-scale 0.9 \
  --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 \
  --max-duration 1200 \
  --master-port 12345

export CUDA_VISIBLE_DEVICES="0"
for m in ctc-decoding attention-decoder-rescoring-no-ngram; do
  ./zipformer/ctc_decode.py \
    --epoch 50 \
    --avg 29 \
    --exp-dir zipformer/exp-large \
    --use-ctc 1 \
    --use-transducer 0 \
    --use-attention-decoder 1 \
    --attention-decoder-loss-scale 0.9 \
    --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 \
    --max-duration 100 \
    --causal 0 \
    --num-paths 100 \
    --decoding-method $m
done

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

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

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

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

export CUDA_VISIBLE_DEVICES="0,1,2,3"
./zipformer/train.py \
  --world-size 4 \
  --num-epochs 50 \
  --start-epoch 1 \
  --use-fp16 0 \
  --use-bf16 1 \
  --exp-dir zipformer/exp_amp_bf16 \
  --causal 0 \
  --full-libri 1 \
  --max-duration 1000

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

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

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

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

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

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

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

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

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

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

./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

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

# 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

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

# 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

./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

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

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

./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

./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

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

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

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

./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

./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

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

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

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

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

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

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

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

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

./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

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

./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

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

./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

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

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

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

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

./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

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

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

#!/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

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

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

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

./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

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

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

./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

./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

./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

./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

./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

./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

./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

./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

./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

./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

./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

./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

./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

./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

./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

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

./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

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

./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

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

./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

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

./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

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

./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

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

./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

./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

./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

./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

./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

./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

./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

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

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

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

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

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

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

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

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

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

./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

./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

./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

./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

./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

./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

# 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

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

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

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

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

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

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

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

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

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

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

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

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

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 \

    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

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

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