mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-09-03 22:24:19 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'dev/aishell-zipformer-bbpe' of https://github.com/jinzr/icefall into dev/aishell-zipformer-bbpe

Browse Source
This commit is contained in:
JinZr 2024-01-11 09:41:40 +08:00
commit 585ae128df
1 changed files with 11 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
egs/aishell/ASR/zipformer_bbpe/decode.py
View File

@ -21,29 +21,29 @@
"""
Usage:
(1) greedy search
./zipformer/decode.py \
./zipformer_bbpe/decode.py \
--epoch 35 \
--avg 15 \
--exp-dir ./zipformer/exp \
--exp-dir ./zipformer_bbpe/exp \
--lang-dir data/lang_char \
--max-duration 600 \
--decoding-method greedy_search
(2) modified beam search
./zipformer/decode.py \
./zipformer_bbpe/decode.py \
--epoch 35 \
--avg 15 \
--exp-dir ./zipformer/exp \
--exp-dir ./zipformer_bbpe/exp \
--lang-dir data/lang_char \
--max-duration 600 \
--decoding-method modified_beam_search \
--beam-size 4
(3) fast beam search (trivial_graph)
./zipformer/decode.py \
./zipformer_bbpe/decode.py \
--epoch 35 \
--avg 15 \
--exp-dir ./zipformer/exp \
--exp-dir ./zipformer_bbpe/exp \
--lang-dir data/lang_char \
--max-duration 600 \
--decoding-method fast_beam_search \
@ -52,10 +52,10 @@ Usage:
--max-states 64
(4) fast beam search (LG)
./zipformer/decode.py \
./zipformer_bbpe/decode.py \
--epoch 30 \
--avg 15 \
--exp-dir ./zipformer/exp \
--exp-dir ./zipformer_bbpe/exp \
--lang-dir data/lang_char \
--max-duration 600 \
--decoding-method fast_beam_search_LG \
@ -64,10 +64,10 @@ Usage:
--max-states 64
(5) fast beam search (nbest oracle WER)
./zipformer/decode.py \
./zipformer_bbpe/decode.py \
--epoch 35 \
--avg 15 \
--exp-dir ./zipformer/exp \
--exp-dir ./zipformer_bbpe/exp \
--lang-dir data/lang_char \
--max-duration 600 \
--decoding-method fast_beam_search_nbest_oracle \
@ -551,7 +551,7 @@ def decode_dataset(
assert len(hyps) == len(texts)
for cut_id, hyp_words, ref_text in zip(cut_ids, hyps, texts):
ref_words = ref_text.split()
this_batch.append((cut_id, ref_text, hyp_words))
this_batch.append((cut_id, ref_words, hyp_words))
results[name].extend(this_batch)