mirror of
https://github.com/k2-fsa/icefall.git
synced 2025-09-03 22:24:19 +00:00
support large-v3
This commit is contained in:
Yuekai Zhang
parent
fa7ad4dc72
commit
2ce09809cd
@ -49,8 +49,8 @@ def compute_fbank_aishell(num_mel_bins: int = 80, perturb_speed: bool = False):
|
||||
|
||||
dataset_parts = (
|
||||
"train",
|
||||
#"dev",
|
||||
#"test",
|
||||
"dev",
|
||||
"test",
|
||||
)
|
||||
prefix = "aishell"
|
||||
suffix = "jsonl.gz"
|
||||
@ -69,7 +69,7 @@ def compute_fbank_aishell(num_mel_bins: int = 80, perturb_speed: bool = False):
|
||||
dataset_parts,
|
||||
)
|
||||
|
||||
extractor = WhisperFbank(WhisperFbankConfig(device='cuda'))
|
||||
extractor = WhisperFbank(WhisperFbankConfig(num_filters=num_mel_bins, device='cuda'))
|
||||
|
||||
with get_executor() as ex: # Initialize the executor only once.
|
||||
for partition, m in manifests.items():
|
||||
|
||||
@ -83,9 +83,9 @@ if [ $stage -le 0 ] && [ $stop_stage -ge 0 ]; then
|
||||
#
|
||||
# ln -sfv /path/to/musan $dl_dir/musan
|
||||
#
|
||||
if [ ! -d $dl_dir/musan ]; then
|
||||
lhotse download musan $dl_dir
|
||||
fi
|
||||
# if [ ! -d $dl_dir/musan ]; then
|
||||
# lhotse download musan $dl_dir
|
||||
# fi
|
||||
fi
|
||||
|
||||
if [ $stage -le 1 ] && [ $stop_stage -ge 1 ]; then
|
||||
@ -99,17 +99,17 @@ if [ $stage -le 1 ] && [ $stop_stage -ge 1 ]; then
|
||||
fi
|
||||
fi
|
||||
|
||||
if [ $stage -le 2 ] && [ $stop_stage -ge 2 ]; then
|
||||
log "Stage 2: Prepare musan manifest"
|
||||
# We assume that you have downloaded the musan corpus
|
||||
# to data/musan
|
||||
if [ ! -f data/manifests/.musan_manifests.done ]; then
|
||||
log "It may take 6 minutes"
|
||||
mkdir -p data/manifests
|
||||
lhotse prepare musan $dl_dir/musan data/manifests
|
||||
touch data/manifests/.musan_manifests.done
|
||||
fi
|
||||
fi
|
||||
# if [ $stage -le 2 ] && [ $stop_stage -ge 2 ]; then
|
||||
# log "Stage 2: Prepare musan manifest"
|
||||
# # We assume that you have downloaded the musan corpus
|
||||
# # to data/musan
|
||||
# if [ ! -f data/manifests/.musan_manifests.done ]; then
|
||||
# log "It may take 6 minutes"
|
||||
# mkdir -p data/manifests
|
||||
# lhotse prepare musan $dl_dir/musan data/manifests
|
||||
# touch data/manifests/.musan_manifests.done
|
||||
# fi
|
||||
# fi
|
||||
|
||||
if [ $stage -le 3 ] && [ $stop_stage -ge 3 ]; then
|
||||
log "Stage 3: Compute fbank for aishell"
|
||||
@ -120,47 +120,56 @@ if [ $stage -le 3 ] && [ $stop_stage -ge 3 ]; then
|
||||
fi
|
||||
fi
|
||||
|
||||
if [ $stage -le 30 ] && [ $stop_stage -ge 30 ]; then
|
||||
# if [ $stage -le 30 ] && [ $stop_stage -ge 30 ]; then
|
||||
# log "Stage 30: Compute whisper fbank for aishell"
|
||||
# if [ ! -f data/fbank/.aishell.done ]; then
|
||||
# mkdir -p data/fbank
|
||||
# ./local/compute_whisper_fbank_aishell.py --perturb-speed True
|
||||
# touch data/fbank/.aishell.done
|
||||
# fi
|
||||
# fi
|
||||
|
||||
if [ $stage -le 300 ] && [ $stop_stage -ge 300 ]; then
|
||||
log "Stage 30: Compute whisper fbank for aishell"
|
||||
if [ ! -f data/fbank/.aishell.done ]; then
|
||||
mkdir -p data/fbank
|
||||
./local/compute_whisper_fbank_aishell.py --perturb-speed True
|
||||
./local/compute_whisper_fbank_aishell.py --perturb-speed True --num-mel-bins 128
|
||||
touch data/fbank/.aishell.done
|
||||
fi
|
||||
fi
|
||||
|
||||
if [ $stage -le 4 ] && [ $stop_stage -ge 4 ]; then
|
||||
log "Stage 4: Compute fbank for musan"
|
||||
if [ ! -f data/fbank/.msuan.done ]; then
|
||||
mkdir -p data/fbank
|
||||
./local/compute_fbank_musan.py
|
||||
touch data/fbank/.msuan.done
|
||||
fi
|
||||
fi
|
||||
# if [ $stage -le 4 ] && [ $stop_stage -ge 4 ]; then
|
||||
# log "Stage 4: Compute fbank for musan"
|
||||
# if [ ! -f data/fbank/.msuan.done ]; then
|
||||
# mkdir -p data/fbank
|
||||
# ./local/compute_fbank_musan.py
|
||||
# touch data/fbank/.msuan.done
|
||||
# fi
|
||||
# fi
|
||||
|
||||
if [ $stage -le 40 ] && [ $stop_stage -ge 40 ]; then
|
||||
log "Stage 4: Compute fbank for musan"
|
||||
if [ ! -f data/fbank/.msuan.done ]; then
|
||||
mkdir -p data/fbank
|
||||
./local/compute_whisper_fbank_musan.py
|
||||
touch data/fbank/.msuan.done
|
||||
fi
|
||||
fi
|
||||
# if [ $stage -le 40 ] && [ $stop_stage -ge 40 ]; then
|
||||
# log "Stage 4: Compute fbank for musan"
|
||||
# if [ ! -f data/fbank/.msuan.done ]; then
|
||||
# mkdir -p data/fbank
|
||||
# ./local/compute_whisper_fbank_musan.py
|
||||
# touch data/fbank/.msuan.done
|
||||
# fi
|
||||
# fi
|
||||
|
||||
lang_phone_dir=data/lang_phone
|
||||
if [ $stage -le 5 ] && [ $stop_stage -ge 5 ]; then
|
||||
log "Stage 5: Prepare phone based lang"
|
||||
mkdir -p $lang_phone_dir
|
||||
# lang_phone_dir=data/lang_phone
|
||||
# if [ $stage -le 5 ] && [ $stop_stage -ge 5 ]; then
|
||||
# log "Stage 5: Prepare phone based lang"
|
||||
# mkdir -p $lang_phone_dir
|
||||
|
||||
(echo '!SIL SIL'; echo '<SPOKEN_NOISE> SPN'; echo '<UNK> SPN'; ) |
|
||||
cat - $dl_dir/aishell/resource_aishell/lexicon.txt |
|
||||
sort | uniq > $lang_phone_dir/lexicon.txt
|
||||
# (echo '!SIL SIL'; echo '<SPOKEN_NOISE> SPN'; echo '<UNK> SPN'; ) |
|
||||
# cat - $dl_dir/aishell/resource_aishell/lexicon.txt |
|
||||
# sort | uniq > $lang_phone_dir/lexicon.txt
|
||||
|
||||
./local/generate_unique_lexicon.py --lang-dir $lang_phone_dir
|
||||
# ./local/generate_unique_lexicon.py --lang-dir $lang_phone_dir
|
||||
|
||||
if [ ! -f $lang_phone_dir/L_disambig.pt ]; then
|
||||
./local/prepare_lang.py --lang-dir $lang_phone_dir
|
||||
fi
|
||||
# if [ ! -f $lang_phone_dir/L_disambig.pt ]; then
|
||||
# ./local/prepare_lang.py --lang-dir $lang_phone_dir
|
||||
# fi
|
||||
|
||||
|
||||
# Train a bigram P for MMI training
|
||||
@ -173,93 +182,93 @@ if [ $stage -le 5 ] && [ $stop_stage -ge 5 ]; then
|
||||
cut -d " " -f 2- > $lang_phone_dir/transcript_words.txt
|
||||
fi
|
||||
|
||||
if [ ! -f $lang_phone_dir/transcript_tokens.txt ]; then
|
||||
./local/convert_transcript_words_to_tokens.py \
|
||||
--lexicon $lang_phone_dir/uniq_lexicon.txt \
|
||||
--transcript $lang_phone_dir/transcript_words.txt \
|
||||
--oov "<UNK>" \
|
||||
> $lang_phone_dir/transcript_tokens.txt
|
||||
fi
|
||||
# if [ ! -f $lang_phone_dir/transcript_tokens.txt ]; then
|
||||
# ./local/convert_transcript_words_to_tokens.py \
|
||||
# --lexicon $lang_phone_dir/uniq_lexicon.txt \
|
||||
# --transcript $lang_phone_dir/transcript_words.txt \
|
||||
# --oov "<UNK>" \
|
||||
# > $lang_phone_dir/transcript_tokens.txt
|
||||
# fi
|
||||
|
||||
if [ ! -f $lang_phone_dir/P.arpa ]; then
|
||||
./shared/make_kn_lm.py \
|
||||
-ngram-order 2 \
|
||||
-text $lang_phone_dir/transcript_tokens.txt \
|
||||
-lm $lang_phone_dir/P.arpa
|
||||
fi
|
||||
# if [ ! -f $lang_phone_dir/P.arpa ]; then
|
||||
# ./shared/make_kn_lm.py \
|
||||
# -ngram-order 2 \
|
||||
# -text $lang_phone_dir/transcript_tokens.txt \
|
||||
# -lm $lang_phone_dir/P.arpa
|
||||
# fi
|
||||
|
||||
if [ ! -f $lang_phone_dir/P.fst.txt ]; then
|
||||
python3 -m kaldilm \
|
||||
--read-symbol-table="$lang_phone_dir/tokens.txt" \
|
||||
--disambig-symbol='#0' \
|
||||
--max-order=2 \
|
||||
$lang_phone_dir/P.arpa > $lang_phone_dir/P.fst.txt
|
||||
fi
|
||||
fi
|
||||
# if [ ! -f $lang_phone_dir/P.fst.txt ]; then
|
||||
# python3 -m kaldilm \
|
||||
# --read-symbol-table="$lang_phone_dir/tokens.txt" \
|
||||
# --disambig-symbol='#0' \
|
||||
# --max-order=2 \
|
||||
# $lang_phone_dir/P.arpa > $lang_phone_dir/P.fst.txt
|
||||
# fi
|
||||
# fi
|
||||
|
||||
lang_char_dir=data/lang_char
|
||||
if [ $stage -le 6 ] && [ $stop_stage -ge 6 ]; then
|
||||
log "Stage 6: Prepare char based lang"
|
||||
mkdir -p $lang_char_dir
|
||||
# We reuse words.txt from phone based lexicon
|
||||
# so that the two can share G.pt later.
|
||||
# lang_char_dir=data/lang_char
|
||||
# if [ $stage -le 6 ] && [ $stop_stage -ge 6 ]; then
|
||||
# log "Stage 6: Prepare char based lang"
|
||||
# mkdir -p $lang_char_dir
|
||||
# # We reuse words.txt from phone based lexicon
|
||||
# # so that the two can share G.pt later.
|
||||
|
||||
# The transcripts in training set, generated in stage 5
|
||||
cp $lang_phone_dir/transcript_words.txt $lang_char_dir/transcript_words.txt
|
||||
# # The transcripts in training set, generated in stage 5
|
||||
# cp $lang_phone_dir/transcript_words.txt $lang_char_dir/transcript_words.txt
|
||||
|
||||
cat $dl_dir/aishell/data_aishell/transcript/aishell_transcript_v0.8.txt |
|
||||
cut -d " " -f 2- > $lang_char_dir/text
|
||||
# cat $dl_dir/aishell/data_aishell/transcript/aishell_transcript_v0.8.txt |
|
||||
# cut -d " " -f 2- > $lang_char_dir/text
|
||||
|
||||
(echo '<eps> 0'; echo '!SIL 1'; echo '<SPOKEN_NOISE> 2'; echo '<UNK> 3';) \
|
||||
> $lang_char_dir/words.txt
|
||||
# (echo '<eps> 0'; echo '!SIL 1'; echo '<SPOKEN_NOISE> 2'; echo '<UNK> 3';) \
|
||||
# > $lang_char_dir/words.txt
|
||||
|
||||
cat $lang_char_dir/text | sed 's/ /\n/g' | sort -u | sed '/^$/d' \
|
||||
| awk '{print $1" "NR+3}' >> $lang_char_dir/words.txt
|
||||
# cat $lang_char_dir/text | sed 's/ /\n/g' | sort -u | sed '/^$/d' \
|
||||
# | awk '{print $1" "NR+3}' >> $lang_char_dir/words.txt
|
||||
|
||||
num_lines=$(< $lang_char_dir/words.txt wc -l)
|
||||
(echo "#0 $num_lines"; echo "<s> $(($num_lines + 1))"; echo "</s> $(($num_lines + 2))";) \
|
||||
>> $lang_char_dir/words.txt
|
||||
# num_lines=$(< $lang_char_dir/words.txt wc -l)
|
||||
# (echo "#0 $num_lines"; echo "<s> $(($num_lines + 1))"; echo "</s> $(($num_lines + 2))";) \
|
||||
# >> $lang_char_dir/words.txt
|
||||
|
||||
if [ ! -f $lang_char_dir/L_disambig.pt ]; then
|
||||
./local/prepare_char.py --lang-dir $lang_char_dir
|
||||
fi
|
||||
fi
|
||||
# if [ ! -f $lang_char_dir/L_disambig.pt ]; then
|
||||
# ./local/prepare_char.py --lang-dir $lang_char_dir
|
||||
# fi
|
||||
# fi
|
||||
|
||||
if [ $stage -le 7 ] && [ $stop_stage -ge 7 ]; then
|
||||
log "Stage 7: Prepare Byte BPE based lang"
|
||||
# if [ $stage -le 7 ] && [ $stop_stage -ge 7 ]; then
|
||||
# log "Stage 7: Prepare Byte BPE based lang"
|
||||
|
||||
for vocab_size in ${vocab_sizes[@]}; do
|
||||
lang_dir=data/lang_bbpe_${vocab_size}
|
||||
mkdir -p $lang_dir
|
||||
# for vocab_size in ${vocab_sizes[@]}; do
|
||||
# lang_dir=data/lang_bbpe_${vocab_size}
|
||||
# mkdir -p $lang_dir
|
||||
|
||||
cp $lang_char_dir/words.txt $lang_dir
|
||||
cp $lang_char_dir/text $lang_dir
|
||||
# cp $lang_char_dir/words.txt $lang_dir
|
||||
# cp $lang_char_dir/text $lang_dir
|
||||
|
||||
if [ ! -f $lang_dir/bbpe.model ]; then
|
||||
./local/train_bbpe_model.py \
|
||||
--lang-dir $lang_dir \
|
||||
--vocab-size $vocab_size \
|
||||
--transcript $lang_dir/text
|
||||
fi
|
||||
# if [ ! -f $lang_dir/bbpe.model ]; then
|
||||
# ./local/train_bbpe_model.py \
|
||||
# --lang-dir $lang_dir \
|
||||
# --vocab-size $vocab_size \
|
||||
# --transcript $lang_dir/text
|
||||
# fi
|
||||
|
||||
if [ ! -f $lang_dir/L_disambig.pt ]; then
|
||||
./local/prepare_lang_bbpe.py --lang-dir $lang_dir
|
||||
fi
|
||||
done
|
||||
fi
|
||||
# if [ ! -f $lang_dir/L_disambig.pt ]; then
|
||||
# ./local/prepare_lang_bbpe.py --lang-dir $lang_dir
|
||||
# fi
|
||||
# done
|
||||
# fi
|
||||
|
||||
if [ $stage -le 8 ] && [ $stop_stage -ge 8 ]; then
|
||||
log "Stage 8: Prepare G"
|
||||
# if [ $stage -le 8 ] && [ $stop_stage -ge 8 ]; then
|
||||
# log "Stage 8: Prepare G"
|
||||
|
||||
mkdir -p data/lm
|
||||
# mkdir -p data/lm
|
||||
|
||||
# Train LM on transcripts
|
||||
if [ ! -f data/lm/3-gram.unpruned.arpa ]; then
|
||||
python3 ./shared/make_kn_lm.py \
|
||||
-ngram-order 3 \
|
||||
-text $lang_char_dir/transcript_words.txt \
|
||||
-lm data/lm/3-gram.unpruned.arpa
|
||||
fi
|
||||
# # Train LM on transcripts
|
||||
# if [ ! -f data/lm/3-gram.unpruned.arpa ]; then
|
||||
# python3 ./shared/make_kn_lm.py \
|
||||
# -ngram-order 3 \
|
||||
# -text $lang_char_dir/transcript_words.txt \
|
||||
# -lm data/lm/3-gram.unpruned.arpa
|
||||
# fi
|
||||
|
||||
# We assume you have installed kaldilm, if not, please install
|
||||
# it using: pip install kaldilm
|
||||
@ -285,112 +294,112 @@ if [ $stage -le 8 ] && [ $stop_stage -ge 8 ]; then
|
||||
fi
|
||||
fi
|
||||
|
||||
if [ $stage -le 9 ] && [ $stop_stage -ge 9 ]; then
|
||||
log "Stage 9: Compile LG & HLG"
|
||||
./local/compile_hlg.py --lang-dir $lang_phone_dir --lm G_3_gram_phone
|
||||
./local/compile_hlg.py --lang-dir $lang_char_dir --lm G_3_gram_char
|
||||
for vocab_size in ${vocab_sizes[@]}; do
|
||||
lang_dir=data/lang_bbpe_${vocab_size}
|
||||
./local/compile_hlg.py --lang-dir $lang_dir --lm G_3_gram_char
|
||||
done
|
||||
# if [ $stage -le 9 ] && [ $stop_stage -ge 9 ]; then
|
||||
# log "Stage 9: Compile LG & HLG"
|
||||
# ./local/compile_hlg.py --lang-dir $lang_phone_dir --lm G_3_gram_phone
|
||||
# ./local/compile_hlg.py --lang-dir $lang_char_dir --lm G_3_gram_char
|
||||
# for vocab_size in ${vocab_sizes[@]}; do
|
||||
# lang_dir=data/lang_bbpe_${vocab_size}
|
||||
# ./local/compile_hlg.py --lang-dir $lang_dir --lm G_3_gram_char
|
||||
# done
|
||||
|
||||
./local/compile_lg.py --lang-dir $lang_phone_dir --lm G_3_gram_phone
|
||||
./local/compile_lg.py --lang-dir $lang_char_dir --lm G_3_gram_char
|
||||
for vocab_size in ${vocab_sizes[@]}; do
|
||||
lang_dir=data/lang_bbpe_${vocab_size}
|
||||
./local/compile_lg.py --lang-dir $lang_dir --lm G_3_gram_char
|
||||
done
|
||||
fi
|
||||
# ./local/compile_lg.py --lang-dir $lang_phone_dir --lm G_3_gram_phone
|
||||
# ./local/compile_lg.py --lang-dir $lang_char_dir --lm G_3_gram_char
|
||||
# for vocab_size in ${vocab_sizes[@]}; do
|
||||
# lang_dir=data/lang_bbpe_${vocab_size}
|
||||
# ./local/compile_lg.py --lang-dir $lang_dir --lm G_3_gram_char
|
||||
# done
|
||||
# fi
|
||||
|
||||
if [ $stage -le 10 ] && [ $stop_stage -ge 10 ]; then
|
||||
log "Stage 10: Generate LM training data"
|
||||
# if [ $stage -le 10 ] && [ $stop_stage -ge 10 ]; then
|
||||
# log "Stage 10: Generate LM training data"
|
||||
|
||||
log "Processing char based data"
|
||||
out_dir=data/lm_training_char
|
||||
mkdir -p $out_dir $dl_dir/lm
|
||||
# log "Processing char based data"
|
||||
# out_dir=data/lm_training_char
|
||||
# mkdir -p $out_dir $dl_dir/lm
|
||||
|
||||
if [ ! -f $dl_dir/lm/aishell-train-word.txt ]; then
|
||||
cp $lang_phone_dir/transcript_words.txt $dl_dir/lm/aishell-train-word.txt
|
||||
fi
|
||||
# if [ ! -f $dl_dir/lm/aishell-train-word.txt ]; then
|
||||
# cp $lang_phone_dir/transcript_words.txt $dl_dir/lm/aishell-train-word.txt
|
||||
# fi
|
||||
|
||||
# training words
|
||||
./local/prepare_char_lm_training_data.py \
|
||||
--lang-char data/lang_char \
|
||||
--lm-data $dl_dir/lm/aishell-train-word.txt \
|
||||
--lm-archive $out_dir/lm_data.pt
|
||||
# # training words
|
||||
# ./local/prepare_char_lm_training_data.py \
|
||||
# --lang-char data/lang_char \
|
||||
# --lm-data $dl_dir/lm/aishell-train-word.txt \
|
||||
# --lm-archive $out_dir/lm_data.pt
|
||||
|
||||
# valid words
|
||||
if [ ! -f $dl_dir/lm/aishell-valid-word.txt ]; then
|
||||
aishell_text=$dl_dir/aishell/data_aishell/transcript/aishell_transcript_v0.8.txt
|
||||
aishell_valid_uid=$dl_dir/aishell/data_aishell/transcript/aishell_valid_uid
|
||||
find $dl_dir/aishell/data_aishell/wav/dev -name "*.wav" | sed 's/\.wav//g' | awk -F '/' '{print $NF}' > $aishell_valid_uid
|
||||
awk 'NR==FNR{uid[$1]=$1} NR!=FNR{if($1 in uid) print $0}' $aishell_valid_uid $aishell_text |
|
||||
cut -d " " -f 2- > $dl_dir/lm/aishell-valid-word.txt
|
||||
fi
|
||||
# # valid words
|
||||
# if [ ! -f $dl_dir/lm/aishell-valid-word.txt ]; then
|
||||
# aishell_text=$dl_dir/aishell/data_aishell/transcript/aishell_transcript_v0.8.txt
|
||||
# aishell_valid_uid=$dl_dir/aishell/data_aishell/transcript/aishell_valid_uid
|
||||
# find $dl_dir/aishell/data_aishell/wav/dev -name "*.wav" | sed 's/\.wav//g' | awk -F '/' '{print $NF}' > $aishell_valid_uid
|
||||
# awk 'NR==FNR{uid[$1]=$1} NR!=FNR{if($1 in uid) print $0}' $aishell_valid_uid $aishell_text |
|
||||
# cut -d " " -f 2- > $dl_dir/lm/aishell-valid-word.txt
|
||||
# fi
|
||||
|
||||
./local/prepare_char_lm_training_data.py \
|
||||
--lang-char data/lang_char \
|
||||
--lm-data $dl_dir/lm/aishell-valid-word.txt \
|
||||
--lm-archive $out_dir/lm_data_valid.pt
|
||||
# ./local/prepare_char_lm_training_data.py \
|
||||
# --lang-char data/lang_char \
|
||||
# --lm-data $dl_dir/lm/aishell-valid-word.txt \
|
||||
# --lm-archive $out_dir/lm_data_valid.pt
|
||||
|
||||
# test words
|
||||
if [ ! -f $dl_dir/lm/aishell-test-word.txt ]; then
|
||||
aishell_text=$dl_dir/aishell/data_aishell/transcript/aishell_transcript_v0.8.txt
|
||||
aishell_test_uid=$dl_dir/aishell/data_aishell/transcript/aishell_test_uid
|
||||
find $dl_dir/aishell/data_aishell/wav/test -name "*.wav" | sed 's/\.wav//g' | awk -F '/' '{print $NF}' > $aishell_test_uid
|
||||
awk 'NR==FNR{uid[$1]=$1} NR!=FNR{if($1 in uid) print $0}' $aishell_test_uid $aishell_text |
|
||||
cut -d " " -f 2- > $dl_dir/lm/aishell-test-word.txt
|
||||
fi
|
||||
# # test words
|
||||
# if [ ! -f $dl_dir/lm/aishell-test-word.txt ]; then
|
||||
# aishell_text=$dl_dir/aishell/data_aishell/transcript/aishell_transcript_v0.8.txt
|
||||
# aishell_test_uid=$dl_dir/aishell/data_aishell/transcript/aishell_test_uid
|
||||
# find $dl_dir/aishell/data_aishell/wav/test -name "*.wav" | sed 's/\.wav//g' | awk -F '/' '{print $NF}' > $aishell_test_uid
|
||||
# awk 'NR==FNR{uid[$1]=$1} NR!=FNR{if($1 in uid) print $0}' $aishell_test_uid $aishell_text |
|
||||
# cut -d " " -f 2- > $dl_dir/lm/aishell-test-word.txt
|
||||
# fi
|
||||
|
||||
./local/prepare_char_lm_training_data.py \
|
||||
--lang-char data/lang_char \
|
||||
--lm-data $dl_dir/lm/aishell-test-word.txt \
|
||||
--lm-archive $out_dir/lm_data_test.pt
|
||||
fi
|
||||
# ./local/prepare_char_lm_training_data.py \
|
||||
# --lang-char data/lang_char \
|
||||
# --lm-data $dl_dir/lm/aishell-test-word.txt \
|
||||
# --lm-archive $out_dir/lm_data_test.pt
|
||||
# fi
|
||||
|
||||
|
||||
if [ $stage -le 11 ] && [ $stop_stage -ge 11 ]; then
|
||||
log "Stage 11: Sort LM training data"
|
||||
# Sort LM training data by sentence length in descending order
|
||||
# for ease of training.
|
||||
#
|
||||
# Sentence length equals to the number of tokens
|
||||
# in a sentence.
|
||||
# if [ $stage -le 11 ] && [ $stop_stage -ge 11 ]; then
|
||||
# log "Stage 11: Sort LM training data"
|
||||
# # Sort LM training data by sentence length in descending order
|
||||
# # for ease of training.
|
||||
# #
|
||||
# # Sentence length equals to the number of tokens
|
||||
# # in a sentence.
|
||||
|
||||
out_dir=data/lm_training_char
|
||||
mkdir -p $out_dir
|
||||
ln -snf ../../../librispeech/ASR/local/sort_lm_training_data.py local/
|
||||
# out_dir=data/lm_training_char
|
||||
# mkdir -p $out_dir
|
||||
# ln -snf ../../../librispeech/ASR/local/sort_lm_training_data.py local/
|
||||
|
||||
./local/sort_lm_training_data.py \
|
||||
--in-lm-data $out_dir/lm_data.pt \
|
||||
--out-lm-data $out_dir/sorted_lm_data.pt \
|
||||
--out-statistics $out_dir/statistics.txt
|
||||
# ./local/sort_lm_training_data.py \
|
||||
# --in-lm-data $out_dir/lm_data.pt \
|
||||
# --out-lm-data $out_dir/sorted_lm_data.pt \
|
||||
# --out-statistics $out_dir/statistics.txt
|
||||
|
||||
./local/sort_lm_training_data.py \
|
||||
--in-lm-data $out_dir/lm_data_valid.pt \
|
||||
--out-lm-data $out_dir/sorted_lm_data-valid.pt \
|
||||
--out-statistics $out_dir/statistics-valid.txt
|
||||
# ./local/sort_lm_training_data.py \
|
||||
# --in-lm-data $out_dir/lm_data_valid.pt \
|
||||
# --out-lm-data $out_dir/sorted_lm_data-valid.pt \
|
||||
# --out-statistics $out_dir/statistics-valid.txt
|
||||
|
||||
./local/sort_lm_training_data.py \
|
||||
--in-lm-data $out_dir/lm_data_test.pt \
|
||||
--out-lm-data $out_dir/sorted_lm_data-test.pt \
|
||||
--out-statistics $out_dir/statistics-test.txt
|
||||
fi
|
||||
# ./local/sort_lm_training_data.py \
|
||||
# --in-lm-data $out_dir/lm_data_test.pt \
|
||||
# --out-lm-data $out_dir/sorted_lm_data-test.pt \
|
||||
# --out-statistics $out_dir/statistics-test.txt
|
||||
# fi
|
||||
|
||||
if [ $stage -le 12 ] && [ $stop_stage -ge 12 ]; then
|
||||
log "Stage 11: Train RNN LM model"
|
||||
python ../../../icefall/rnn_lm/train.py \
|
||||
--start-epoch 0 \
|
||||
--world-size 1 \
|
||||
--num-epochs 20 \
|
||||
--use-fp16 0 \
|
||||
--embedding-dim 512 \
|
||||
--hidden-dim 512 \
|
||||
--num-layers 2 \
|
||||
--batch-size 400 \
|
||||
--exp-dir rnnlm_char/exp \
|
||||
--lm-data $out_dir/sorted_lm_data.pt \
|
||||
--lm-data-valid $out_dir/sorted_lm_data-valid.pt \
|
||||
--vocab-size 4336 \
|
||||
--master-port 12345
|
||||
fi
|
||||
# if [ $stage -le 12 ] && [ $stop_stage -ge 12 ]; then
|
||||
# log "Stage 11: Train RNN LM model"
|
||||
# python ../../../icefall/rnn_lm/train.py \
|
||||
# --start-epoch 0 \
|
||||
# --world-size 1 \
|
||||
# --num-epochs 20 \
|
||||
# --use-fp16 0 \
|
||||
# --embedding-dim 512 \
|
||||
# --hidden-dim 512 \
|
||||
# --num-layers 2 \
|
||||
# --batch-size 400 \
|
||||
# --exp-dir rnnlm_char/exp \
|
||||
# --lm-data $out_dir/sorted_lm_data.pt \
|
||||
# --lm-data-valid $out_dir/sorted_lm_data-valid.pt \
|
||||
# --vocab-size 4336 \
|
||||
# --master-port 12345
|
||||
# fi
|
||||
|
||||
@ -176,7 +176,7 @@ class AishellAsrDataModule:
|
||||
group.add_argument(
|
||||
"--enable-musan",
|
||||
type=str2bool,
|
||||
default=True,
|
||||
default=False,
|
||||
help="When enabled, select noise from MUSAN and mix it"
|
||||
"with training dataset. ",
|
||||
)
|
||||
@ -192,11 +192,11 @@ class AishellAsrDataModule:
|
||||
The state dict for the training sampler.
|
||||
"""
|
||||
logging.info("About to get Musan cuts")
|
||||
cuts_musan = load_manifest(self.args.manifest_dir / "musan_cuts.jsonl.gz")
|
||||
|
||||
transforms = []
|
||||
if self.args.enable_musan:
|
||||
logging.info("Enable MUSAN")
|
||||
cuts_musan = load_manifest(self.args.manifest_dir / "musan_cuts.jsonl.gz")
|
||||
transforms.append(
|
||||
CutMix(cuts=cuts_musan, p=0.5, snr=(10, 20), preserve_id=True)
|
||||
)
|
||||
|
||||
@ -127,6 +127,15 @@ def get_parser():
|
||||
help="The experiment dir",
|
||||
)
|
||||
|
||||
parser.add_argument(
|
||||
"--model-name",
|
||||
type=str,
|
||||
default="large-v2",
|
||||
choices=["large-v2", "large-v3", "medium", "small", "tiny"],
|
||||
help="""The model name to use.
|
||||
""",
|
||||
)
|
||||
|
||||
return parser
|
||||
|
||||
|
||||
@ -370,7 +379,7 @@ def main():
|
||||
|
||||
logging.info(f"device: {device}")
|
||||
|
||||
model = whisper.load_model("medium")
|
||||
model = whisper.load_model(params.model_name)
|
||||
if params.epoch > 0:
|
||||
if params.avg > 1:
|
||||
start = params.epoch - params.avg
|
||||
|
||||
@ -2,10 +2,10 @@
|
||||
"fp16": {
|
||||
"enabled": true,
|
||||
"loss_scale": 0,
|
||||
"loss_scale_window": 1000,
|
||||
"loss_scale_window": 100,
|
||||
"initial_scale_power": 16,
|
||||
"hysteresis": 2,
|
||||
"min_loss_scale": 1
|
||||
"min_loss_scale": 0.01
|
||||
},
|
||||
"zero_optimization": {
|
||||
"stage": 1,
|
||||
@ -19,8 +19,8 @@
|
||||
"scheduler": {
|
||||
"type": "WarmupLR",
|
||||
"params": {
|
||||
"warmup_min_lr": 5e-6,
|
||||
"warmup_max_lr": 1e-5,
|
||||
"warmup_min_lr": 1e-6,
|
||||
"warmup_max_lr": 5e-6,
|
||||
"warmup_num_steps": 100
|
||||
}
|
||||
},
|
||||
|
||||
@ -2,6 +2,8 @@ import torch
|
||||
import torch.nn as nn
|
||||
import base64
|
||||
import gzip
|
||||
import warnings
|
||||
from tqdm import tqdm
|
||||
from dataclasses import dataclass
|
||||
from typing import Dict, Iterable, Optional, Union
|
||||
import os
|
||||
@ -275,6 +277,11 @@ class Whisper(nn.Module):
|
||||
@property
|
||||
def is_multilingual(self):
|
||||
return self.dims.n_vocab == 51865
|
||||
return self.dims.n_vocab >= 51865
|
||||
|
||||
@property
|
||||
def num_languages(self):
|
||||
return self.dims.n_vocab - 51765 - int(self.is_multilingual)
|
||||
|
||||
def install_kv_cache_hooks(self, cache: Optional[dict] = None):
|
||||
"""
|
||||
@ -324,6 +331,7 @@ _MODELS = {
|
||||
"medium": "https://openaipublic.azureedge.net/main/whisper/models/345ae4da62f9b3d59415adc60127b97c714f32e89e936602e85993674d08dcb1/medium.pt",
|
||||
"large-v1": "https://openaipublic.azureedge.net/main/whisper/models/e4b87e7e0bf463eb8e6956e646f1e277e901512310def2c24bf0e11bd3c28e9a/large-v1.pt",
|
||||
"large-v2": "https://openaipublic.azureedge.net/main/whisper/models/81f7c96c852ee8fc832187b0132e569d6c3065a3252ed18e56effd0b6a73e524/large-v2.pt",
|
||||
"large-v3": "https://openaipublic.azureedge.net/main/whisper/models/e5b1a55b89c1367dacf97e3e19bfd829a01529dbfdeefa8caeb59b3f1b81dadb/large-v3.pt",
|
||||
"large": "https://openaipublic.azureedge.net/main/whisper/models/81f7c96c852ee8fc832187b0132e569d6c3065a3252ed18e56effd0b6a73e524/large-v2.pt",
|
||||
}
|
||||
|
||||
|
||||
@ -1,11 +1,11 @@
|
||||
k2
|
||||
kaldialign
|
||||
lhotse==1.18
|
||||
#git+https://github.com/lhotse-speech/lhotse
|
||||
#lhotse==1.18
|
||||
git+https://github.com/lhotse-speech/lhotse
|
||||
sentencepiece
|
||||
tensorboard
|
||||
librosa
|
||||
openai-whisper
|
||||
openai-whisper==20231117
|
||||
zhconv
|
||||
WeTextProcessing
|
||||
deepspeed
|
||||
|
||||
@ -796,7 +796,7 @@ def run(rank, world_size, args):
|
||||
logging.info(f"Number of model parameters: {num_param}")
|
||||
|
||||
tokenizer = whisper.tokenizer.get_tokenizer(
|
||||
model.is_multilingual, language="zh", task="transcribe"
|
||||
model.is_multilingual, num_languages=model.num_languages, language="zh", task="transcribe"
|
||||
)
|
||||
|
||||
assert params.save_every_n >= params.average_period
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user