This commit is contained in:
AmirHussein96 2024-09-24 03:07:37 +00:00
parent 64c1413e76
commit 99700dc71b

View File

@ -39,19 +39,43 @@ export CUDA_VISIBLE_DEVICES="0,1,2,3"
The decoding command is: The decoding command is:
```bash ```bash
export CUDA_VISIBLE_DEVICES="0" export CUDA_VISIBLE_DEVICES="0"
for m in ctc-decoding 1best nbest nbest-rescoring whole-lattice-rescoring; do for method in modified_beam_search greedy_search; do
./zipformer/ctc_decode.py \ ./zipformer_hat/decode.py \
--epoch 40 \ --epoch 40 --avg 16 --use-averaged-model True \
--avg 16 \ --beam-size 4 \
--exp-dir zipformer/exp-ctc-rnnt \ --exp-dir ./zipformer_hat/exp \
--use-transducer 1 \ --bpe-model data/lang_bpe_500/bpe.model \
--use-ctc 1 \ --max-contexts 4 \
--max-duration 300 \ --max-states 8 \
--causal 0 \ --max-duration 600 \
--num-paths 100 \ --decoding-method $method
--nbest-scale 1.0 \ done
--hlg-scale 0.6 \ ```
--decoding-method $m
The decoding with shallow LM fusion and ILM subtraction:
```bash
for method in modified_beam_search_auxlm_shallow_fusion; do
./zipformer_hat/decode.py \
--epoch 40 --avg 16 --use-averaged-model True \
--beam-size 4 \
--exp-dir ./zipformer_hat/exp \
--bpe-model data/lang_bpe_500/bpe.model \
--max-contexts 4 \
--max-states 8 \
--max-duration 800 \
--decoding-method $method \
--subtract-ilm True\
--ilm-scale 0.1 \
--use-shallow-fusion 1 \
--lm-type rnn \
--lm-exp-dir rnn_lm/exp \
--lm-epoch 25 \
--lm-scale 0.45 \
--lm-avg 5 \
--lm-vocab-size 500 \
--rnn-lm-embedding-dim 512 \
--rnn-lm-hidden-dim 512 \
--rnn-lm-num-layers 2
done done
``` ```