speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #525 from alibaba-damo-academy/dev_lyh

Browse Source 
fix m2met2 baseline bugs
This commit is contained in:
yhliang 2023-05-18 11:41:44 +08:00 committed by GitHub
commit 0ec7f0aea6
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
5 changed files with 15 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
egs/alimeeting/sa-asr/asr_local.sh
View File

@ -1153,10 +1153,10 @@ if ! "${skip_train}"; then
mkdir -p ${sa_asr_exp}/log
INIT_FILE=${sa_asr_exp}/ddp_init
if [ ! -f "exp/damo/speech_xvector_sv-zh-cn-cnceleb-16k-spk3465-pytorch/sv.pth" ]; then
if [ ! -f "exp/damo/speech_xvector_sv-zh-cn-cnceleb-16k-spk3465-pytorch/sv.pb" ]; then
# download xvector extractor model file
python local/download_xvector_model.py exp
log "Successfully download the pretrained xvector extractor to exp/damo/speech_xvector_sv-zh-cn-cnceleb-16k-spk3465-pytorch/sv.pth"
log "Successfully download the pretrained xvector extractor to exp/damo/speech_xvector_sv-zh-cn-cnceleb-16k-spk3465-pytorch/sv.pb"
fi
if [ -f $INIT_FILE ];then
@ -1195,8 +1195,8 @@ if ! "${skip_train}"; then
--init_param "${asr_exp}/valid.acc.ave.pb:decoder.decoders.3:decoder.decoder4.2" \
--init_param "${asr_exp}/valid.acc.ave.pb:decoder.decoders.4:decoder.decoder4.3" \
--init_param "${asr_exp}/valid.acc.ave.pb:decoder.decoders.5:decoder.decoder4.4" \
--init_param "exp/damo/speech_xvector_sv-zh-cn-cnceleb-16k-spk3465-pytorch/sv.pth:encoder:spk_encoder" \
--init_param "exp/damo/speech_xvector_sv-zh-cn-cnceleb-16k-spk3465-pytorch/sv.pth:decoder:spk_encoder:decoder.output_dense" \
--init_param "exp/damo/speech_xvector_sv-zh-cn-cnceleb-16k-spk3465-pytorch/sv.pb:encoder:spk_encoder" \
--init_param "exp/damo/speech_xvector_sv-zh-cn-cnceleb-16k-spk3465-pytorch/sv.pb:decoder:spk_encoder:decoder.output_dense" \
--valid_data_path_and_name_and_type "${_asr_valid_dir}/${_scp},speech,${_type}" \
--valid_data_path_and_name_and_type "${_asr_valid_dir}/text,text,text" \
--valid_data_path_and_name_and_type "${_asr_valid_dir}/oracle_profile_nopadding.scp,profile,npy" \

11
egs/alimeeting/sa-asr/local/alimeeting_data_prep.sh
View File

@ -61,9 +61,9 @@ mkdir -p $near_dir
if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
log "stage 1:process alimeeting near dir"
find -L $near_raw_dir/audio_dir -iname "*.wav" > $near_dir/wavlist
find -L $near_raw_dir/audio_dir -iname "*.wav" | sort > $near_dir/wavlist
awk -F '/' '{print $NF}' $near_dir/wavlist | awk -F '.' '{print $1}' > $near_dir/uttid
find -L $near_raw_dir/textgrid_dir -iname "*.TextGrid" > $near_dir/textgrid.flist
find -L $near_raw_dir/textgrid_dir -iname "*.TextGrid" | sort > $near_dir/textgrid.flist
n1_wav=$(wc -l < $near_dir/wavlist)
n2_text=$(wc -l < $near_dir/textgrid.flist)
log near file found $n1_wav wav and $n2_text text.
@ -90,9 +90,9 @@ fi
if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
log "stage 2:process alimeeting far dir"
find -L $far_raw_dir/audio_dir -iname "*.wav" > $far_dir/wavlist
find -L $far_raw_dir/audio_dir -iname "*.wav" | sort > $far_dir/wavlist
awk -F '/' '{print $NF}' $far_dir/wavlist | awk -F '.' '{print $1}' > $far_dir/uttid
find -L $far_raw_dir/textgrid_dir -iname "*.TextGrid" > $far_dir/textgrid.flist
find -L $far_raw_dir/textgrid_dir -iname "*.TextGrid" | sort > $far_dir/textgrid.flist
n1_wav=$(wc -l < $far_dir/wavlist)
n2_text=$(wc -l < $far_dir/textgrid.flist)
log far file found $n1_wav wav and $n2_text text.
@ -120,7 +120,8 @@ fi
if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
log "stage 3: finali data process"
local/fix_data_dir.sh $near_dir
local/fix_data_dir.sh $far_dir
local/copy_data_dir.sh $near_dir data/${tgt}_Ali_near
local/copy_data_dir.sh $far_dir data/${tgt}_Ali_far

4
egs/alimeeting/sa-asr/local/gen_oracle_profile_padding.py
View File

@ -42,8 +42,8 @@ if __name__=="__main__":
global_spk_list_tmp = global_spk_list[: ]
for spk in meeting_map_tmp[meeting]:
global_spk_list_tmp.remove(spk)
padding_spk = random.sample(global_spk_list_tmp, 4 - num)
meeting_map_tmp[meeting] = meeting_map_tmp[meeting] + padding_spk
padding_spk = random.sample(global_spk_list_tmp, 4 - num)
meeting_map_tmp[meeting] = meeting_map_tmp[meeting] + padding_spk
meeting_map = {}
os.system('mkdir -p ' + path + '/oracle_profile_padding')

2
egs/alimeeting/sa-asr/path.sh
View File

@ -2,4 +2,4 @@ export FUNASR_DIR=$PWD/../../..
# NOTE(kan-bayashi): Use UTF-8 in Python to avoid UnicodeDecodeError when LC_ALL=C
export PYTHONIOENCODING=UTF-8
export PATH=$FUNASR_DIR/funasr/bin:$PATH
export PATH=$FUNASR_DIR/funasr/bin:./utils:$PATH

4
funasr/models/frontend/default.py
View File

@ -102,8 +102,8 @@ class DefaultFrontend(AbsFrontend):
if input_stft.dim() == 4:
# h: (B, T, C, F) -> h: (B, T, F)
if self.training:
if self.use_channel == None:
input_stft = input_stft[:, :, 0, :]
if self.use_channel is not None:
input_stft = input_stft[:, :, self.use_channel, :]
else:
# Select 1ch randomly
ch = np.random.randint(input_stft.size(2))