deepgeek/CosyVoice
2
0
Fork 0
mirror of https://github.com/FunAudioLLM/CosyVoice.git synced 2026-02-05 18:09:24 +08:00
Code Issues Packages Projects Releases Wiki Activity

online feature

Browse Source
This commit is contained in:
lyuxiang.lx
2026-01-28 15:19:07 +00:00
parent 1822c5c908
commit 66b80dbccb
22 changed files with 133 additions and 116 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
examples/libritts/cosyvoice2/run.sh
View File

@@ -24,27 +24,12 @@ if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
done
fi
if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
echo "Extract campplus speaker embedding, you will get spk2embedding.pt and utt2embedding.pt in data/$x dir"
for x in train-clean-100 train-clean-360 train-other-500 dev-clean dev-other test-clean test-other; do
tools/extract_embedding.py --dir data/$x \
--onnx_path $pretrained_model_dir/campplus.onnx
done
fi
if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
echo "Extract discrete speech token, you will get utt2speech_token.pt in data/$x dir"
for x in train-clean-100 train-clean-360 train-other-500 dev-clean dev-other test-clean test-other; do
tools/extract_speech_token.py --dir data/$x \
--onnx_path $pretrained_model_dir/speech_tokenizer_v2.onnx
done
fi
# NOTE embedding/token extraction is not necessary now as we support online feature extraction
if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
echo "Prepare required parquet format data, you should have prepared wav.scp/text/utt2spk/spk2utt/utt2embedding.pt/spk2embedding.pt/utt2speech_token.pt"
for x in train-clean-100 train-clean-360 train-other-500 dev-clean dev-other test-clean test-other; do
mkdir -p data/$x/parquet
tools/make_parquet_list.py --num_utts_per_parquet 1000 \
../../../tools/make_parquet_list.py --num_utts_per_parquet 1000 \
--num_processes 10 \
--src_dir data/$x \
--des_dir data/$x/parquet
@@ -69,12 +54,13 @@ if [ ${stage} -le 5 ] && [ ${stop_stage} -ge 5 ]; then
for model in llm flow hifigan; do
torchrun --nnodes=1 --nproc_per_node=$num_gpus \
--rdzv_id=$job_id --rdzv_backend="c10d" --rdzv_endpoint="localhost:1234" \
cosyvoice/bin/train.py \
../../../cosyvoice/bin/train.py \
--train_engine $train_engine \
--config conf/cosyvoice2.yaml \
--train_data data/train.data.list \
--cv_data data/dev.data.list \
--qwen_pretrain_path $pretrained_model_dir/CosyVoice-BlankEN \
--onnx_path $pretrained_model_dir \
--model $model \
--checkpoint $pretrained_model_dir/$model.pt \
--model_dir `pwd`/exp/cosyvoice2/$model/$train_engine \