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

update inference config

Browse Source
This commit is contained in:
lzr265946 2022-12-23 19:05:32 +08:00
parent d653df71cb
commit 532c56b3c1
9 changed files with 25 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
egs_modelscope/aishell/paraformer/paraformer_large_finetune.sh
View File

@ -7,7 +7,7 @@ CUDA_VISIBLE_DEVICES="0,1" # set gpus, e.g., CUDA_VISIBLE_DEVICES="0,1"
gpu_num=2 gpu_num=2
count=1 count=1
gpu_inference=true # Whether to perform gpu decoding, set false for cpu decoding gpu_inference=true # Whether to perform gpu decoding, set false for cpu decoding
njob=4 # the number of jobs for each gpu njob=1 # the number of jobs for each gpu
train_cmd=utils/run.pl train_cmd=utils/run.pl
infer_cmd=utils/run.pl infer_cmd=utils/run.pl
@ -84,7 +84,7 @@ gpuid_list=$CUDA_VISIBLE_DEVICES # set gpus for decoding, the same as training
ngpu=$(echo $gpuid_list | awk -F "," '{print NF}') ngpu=$(echo $gpuid_list | awk -F "," '{print NF}')
if ${gpu_inference}; then if ${gpu_inference}; then
inference_nj=$[${ngpu}*${njob}] inference_nj=$njob
_ngpu=1 _ngpu=1
else else
inference_nj=$njob inference_nj=$njob
@ -237,10 +237,10 @@ if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
fi fi
${infer_cmd} --gpu "${_ngpu}" --max-jobs-run "${_nj}" JOB=1:"${_nj}" "${_logdir}"/asr_inference.JOB.log \ ${infer_cmd} --gpu "${_ngpu}" --max-jobs-run "${_nj}" JOB=1:"${_nj}" "${_logdir}"/asr_inference.JOB.log \
python -m funasr.bin.asr_inference_launch \ python -m funasr.bin.asr_inference_launch \
--batch_size 1 \ --batch_size 64 \
--ngpu "${_ngpu}" \ --ngpu "${_ngpu}" \
--njob ${njob} \ --njob ${njob} \
--gpuid_list ${gpuid_list} \ --gpuid_list ${gpuid_list:0:1} \
--data_path_and_name_and_type "${_data}/${scp},speech,${type}" \ --data_path_and_name_and_type "${_data}/${scp},speech,${type}" \
--key_file "${_logdir}"/keys.JOB.scp \ --key_file "${_logdir}"/keys.JOB.scp \
--asr_train_config "${asr_exp}"/config.yaml \ --asr_train_config "${asr_exp}"/config.yaml \

4
egs_modelscope/aishell/paraformer/paraformer_large_infer.sh
View File

@ -10,9 +10,9 @@ exp_dir=
model_name=speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch model_name=speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch
model_revision="v1.0.4" # please do not modify the model revision model_revision="v1.0.4" # please do not modify the model revision
inference_nj=32 inference_nj=32
gpuid_list="0,1" # set gpus, e.g., gpuid_list="0,1" gpuid_list="0" # set gpus, e.g., gpuid_list="0,1"
ngpu=$(echo $gpuid_list | awk -F "," '{print NF}') ngpu=$(echo $gpuid_list | awk -F "," '{print NF}')
njob=4 # the number of jobs for each gpu njob=1 # the number of jobs for each gpu
gpu_inference=true # Whether to perform gpu decoding, set false for cpu decoding gpu_inference=true # Whether to perform gpu decoding, set false for cpu decoding
if ${gpu_inference}; then if ${gpu_inference}; then

8
egs_modelscope/aishell2/paraformer/paraformer_large_finetune.sh
View File

@ -7,7 +7,7 @@ CUDA_VISIBLE_DEVICES="0,1" # set gpus, e.g., CUDA_VISIBLE_DEVICES="0,1"
gpu_num=2 gpu_num=2
count=1 count=1
gpu_inference=true # Whether to perform gpu decoding, set false for cpu decoding gpu_inference=true # Whether to perform gpu decoding, set false for cpu decoding
njob=4 # the number of jobs for each gpu njob=1 # the number of jobs for each gpu
train_cmd=utils/run.pl train_cmd=utils/run.pl
infer_cmd=utils/run.pl infer_cmd=utils/run.pl
@ -85,7 +85,7 @@ gpuid_list=$CUDA_VISIBLE_DEVICES # set gpus for decoding, the same as training
ngpu=$(echo $gpuid_list | awk -F "," '{print NF}') ngpu=$(echo $gpuid_list | awk -F "," '{print NF}')
if ${gpu_inference}; then if ${gpu_inference}; then
inference_nj=$[${ngpu}*${njob}] inference_nj=$njob
_ngpu=1 _ngpu=1
else else
inference_nj=$njob inference_nj=$njob
@ -252,10 +252,10 @@ if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
fi fi
${infer_cmd} --gpu "${_ngpu}" --max-jobs-run "${_nj}" JOB=1:"${_nj}" "${_logdir}"/asr_inference.JOB.log \ ${infer_cmd} --gpu "${_ngpu}" --max-jobs-run "${_nj}" JOB=1:"${_nj}" "${_logdir}"/asr_inference.JOB.log \
python -m funasr.bin.asr_inference_launch \ python -m funasr.bin.asr_inference_launch \
--batch_size 1 \ --batch_size 64 \
--ngpu "${_ngpu}" \ --ngpu "${_ngpu}" \
--njob ${njob} \ --njob ${njob} \
--gpuid_list ${gpuid_list} \ --gpuid_list ${gpuid_list:0:1} \
--data_path_and_name_and_type "${_data}/${scp},speech,${type}" \ --data_path_and_name_and_type "${_data}/${scp},speech,${type}" \
--key_file "${_logdir}"/keys.JOB.scp \ --key_file "${_logdir}"/keys.JOB.scp \
--asr_train_config "${asr_exp}"/config.yaml \ --asr_train_config "${asr_exp}"/config.yaml \

4
egs_modelscope/aishell2/paraformer/paraformer_large_infer.sh
View File

@ -10,9 +10,9 @@ exp_dir=
model_name=speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch model_name=speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch
model_revision="v1.0.4" # please do not modify the model revision model_revision="v1.0.4" # please do not modify the model revision
inference_nj=32 inference_nj=32
gpuid_list="0,1" # set gpus, e.g., gpuid_list="0,1" gpuid_list="0" # set gpus, e.g., gpuid_list="0,1"
ngpu=$(echo $gpuid_list | awk -F "," '{print NF}') ngpu=$(echo $gpuid_list | awk -F "," '{print NF}')
njob=4 # the number of jobs for each gpu njob=1 # the number of jobs for each gpu
gpu_inference=true # Whether to perform gpu decoding, set false for cpu decoding gpu_inference=true # Whether to perform gpu decoding, set false for cpu decoding
if ${gpu_inference}; then if ${gpu_inference}; then

8
egs_modelscope/common/modelscope_common_finetune.sh
View File

@ -7,7 +7,7 @@ CUDA_VISIBLE_DEVICES="0,1" # set gpus, e.g., CUDA_VISIBLE_DEVICES="0,1"
gpu_num=2 gpu_num=2
count=1 count=1
gpu_inference=true # Whether to perform gpu decoding, set false for cpu decoding gpu_inference=true # Whether to perform gpu decoding, set false for cpu decoding
njob=4 # the number of jobs for each gpu njob=1 # the number of jobs for each gpu
train_cmd=utils/run.pl train_cmd=utils/run.pl
infer_cmd=utils/run.pl infer_cmd=utils/run.pl
@ -84,7 +84,7 @@ gpuid_list=$CUDA_VISIBLE_DEVICES # set gpus for decoding, the same as training
ngpu=$(echo $gpuid_list | awk -F "," '{print NF}') ngpu=$(echo $gpuid_list | awk -F "," '{print NF}')
if ${gpu_inference}; then if ${gpu_inference}; then
inference_nj=$[${ngpu}*${njob}] inference_nj=$njob
_ngpu=1 _ngpu=1
else else
inference_nj=$njob inference_nj=$njob
@ -244,10 +244,10 @@ if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
fi fi
${infer_cmd} --gpu "${_ngpu}" --max-jobs-run "${_nj}" JOB=1:"${_nj}" "${_logdir}"/asr_inference.JOB.log \ ${infer_cmd} --gpu "${_ngpu}" --max-jobs-run "${_nj}" JOB=1:"${_nj}" "${_logdir}"/asr_inference.JOB.log \
python -m funasr.bin.asr_inference_launch \ python -m funasr.bin.asr_inference_launch \
--batch_size 1 \ --batch_size 64 \
--ngpu "${_ngpu}" \ --ngpu "${_ngpu}" \
--njob ${njob} \ --njob ${njob} \
--gpuid_list ${gpuid_list} \ --gpuid_list ${gpuid_list:0:1} \
--data_path_and_name_and_type "${_data}/${scp},speech,${type}" \ --data_path_and_name_and_type "${_data}/${scp},speech,${type}" \
--key_file "${_logdir}"/keys.JOB.scp \ --key_file "${_logdir}"/keys.JOB.scp \
--asr_train_config "${asr_exp}"/config.yaml \ --asr_train_config "${asr_exp}"/config.yaml \

4
egs_modelscope/common/modelscope_common_infer.sh
View File

@ -8,9 +8,9 @@ model_name=speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch #
model_revision="v1.0.4" # please do not modify the model revision model_revision="v1.0.4" # please do not modify the model revision
data_dir= # wav list, ${data_dir}/wav.scp data_dir= # wav list, ${data_dir}/wav.scp
exp_dir="exp" exp_dir="exp"
gpuid_list="0,1" gpuid_list="0"
ngpu=$(echo $gpuid_list | awk -F "," '{print NF}') ngpu=$(echo $gpuid_list | awk -F "," '{print NF}')
njob=4 njob=1
gpu_inference=true gpu_inference=true
decode_cmd=utils/run.pl decode_cmd=utils/run.pl

7
egs_modelscope/common/modelscope_common_infer_after_finetune.sh
View File

@ -8,9 +8,9 @@ pretrained_model_name=speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404
data_dir= # wav list, ${data_dir}/wav.scp data_dir= # wav list, ${data_dir}/wav.scp
finetune_model_name= # fine-tuning model name finetune_model_name= # fine-tuning model name
finetune_exp_dir= # fine-tuning model experiment result path finetune_exp_dir= # fine-tuning model experiment result path
gpuid_list="0,1" gpuid_list="0"
ngpu=$(echo $gpuid_list | awk -F "," '{print NF}') ngpu=$(echo $gpuid_list | awk -F "," '{print NF}')
njob=4 njob=1
gpu_inference=true gpu_inference=true
decode_cmd=utils/run.pl decode_cmd=utils/run.pl
@ -20,7 +20,6 @@ if ${gpu_inference}; then
inference_nj=$[${ngpu}*${njob}] inference_nj=$[${ngpu}*${njob}]
_ngpu=1 _ngpu=1
else else
inference_nj=${njob}
inference_nj=${njob} inference_nj=${njob}
_ngpu=0 _ngpu=0
fi fi
@ -63,4 +62,4 @@ ${decode_cmd} --max-jobs-run "${inference_nj}" JOB=1:"${inference_nj}" "${_logdi
for i in $(seq ${inference_nj}); do for i in $(seq ${inference_nj}); do
cat ${_logdir}/text.${i} cat ${_logdir}/text.${i}
done | sort -k1 >${_dir}/text done | sort -k1 >${_dir}/text

4
egs_modelscope/speechio/paraformer/paraformer_large_infer.sh
View File

@ -10,9 +10,9 @@ exp_dir=
model_name=speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch model_name=speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch
model_revision="v1.0.4" # please do not modify the model revision model_revision="v1.0.4" # please do not modify the model revision
inference_nj=32 inference_nj=32
gpuid_list="0,1" # set gpus, e.g., gpuid_list="0,1" gpuid_list="0" # set gpus, e.g., gpuid_list="0,1"
ngpu=$(echo $gpuid_list | awk -F "," '{print NF}') ngpu=$(echo $gpuid_list | awk -F "," '{print NF}')
njob=4 # the number of jobs for each gpu njob=1 # the number of jobs for each gpu
gpu_inference=true # Whether to perform gpu decoding, set false for cpu decoding gpu_inference=true # Whether to perform gpu decoding, set false for cpu decoding
if ${gpu_inference}; then if ${gpu_inference}; then

4
egs_modelscope/wenetspeech/paraformer/paraformer_large_infer.sh
View File

@ -10,9 +10,9 @@ exp_dir=
model_name=speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch model_name=speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch
model_revision="v1.0.4" # please do not modify the model revision model_revision="v1.0.4" # please do not modify the model revision
inference_nj=32 inference_nj=32
gpuid_list="0,1" # set gpus, e.g., gpuid_list="0,1" gpuid_list="0" # set gpus, e.g., gpuid_list="0,1"
ngpu=$(echo $gpuid_list | awk -F "," '{print NF}') ngpu=$(echo $gpuid_list | awk -F "," '{print NF}')
njob=4 # the number of jobs for each gpu njob=1 # the number of jobs for each gpu
gpu_inference=true # Whether to perform gpu decoding, set false for cpu decoding gpu_inference=true # Whether to perform gpu decoding, set false for cpu decoding
if ${gpu_inference}; then if ${gpu_inference}; then