deepgeek/CosyVoice
2
0
Fork 0
mirror of https://github.com/FunAudioLLM/CosyVoice.git synced 2026-02-04 09:29:25 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix vocoder train

Browse Source
This commit is contained in:
lyuxiang.lx
2025-03-07 16:39:13 +08:00
parent fcc054f64e
commit a69b7e275d
12 changed files with 108 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples/libritts/cosyvoice/conf/cosyvoice.fromscratch.yaml
View File

@@ -147,7 +147,7 @@ hifigan: !new:cosyvoice.hifigan.hifigan.HiFiGan
generator: !ref <hift>
discriminator: !new:cosyvoice.hifigan.discriminator.MultipleDiscriminator
mpd: !new:matcha.hifigan.models.MultiPeriodDiscriminator
mrd: !new:cosyvoice.hifigan.discriminator.MultiResolutionDiscriminator
mrd: !new:cosyvoice.hifigan.discriminator.MultiResSpecDiscriminator
mel_spec_transform: [
!ref <mel_spec_transform1>
]

2
examples/libritts/cosyvoice/conf/cosyvoice.yaml
View File

@@ -147,7 +147,7 @@ hifigan: !new:cosyvoice.hifigan.hifigan.HiFiGan
generator: !ref <hift>
discriminator: !new:cosyvoice.hifigan.discriminator.MultipleDiscriminator
mpd: !new:matcha.hifigan.models.MultiPeriodDiscriminator
mrd: !new:cosyvoice.hifigan.discriminator.MultiResolutionDiscriminator
mrd: !new:cosyvoice.hifigan.discriminator.MultiResSpecDiscriminator
mel_spec_transform: [
!ref <mel_spec_transform1>
]

12
examples/libritts/cosyvoice2/conf/cosyvoice2.yaml
View File

@@ -14,8 +14,8 @@ token_frame_rate: 25
token_mel_ratio: 2
# stream related params
chunk_size: 1 # streaming inference chunk size, in second
num_decoding_left_chunks: 2 # streaming inference flow decoder left chunk size, in second
chunk_size: 2 # streaming inference chunk size, in second
num_decoding_left_chunks: 1 # streaming inference flow decoder left chunk size
# model params
# for all class/function included in this repo, we use !<name> or !<new> for intialization, so that user may find all corresponding class/function according to one single yaml.
@@ -112,11 +112,11 @@ hift: !new:cosyvoice.hifigan.generator.HiFTGenerator
# gan related module
mel_spec_transform1: !name:matcha.utils.audio.mel_spectrogram
n_fft: 1024
n_fft: 1920
num_mels: 80
sampling_rate: !ref <sample_rate>
hop_size: 256
win_size: 1024
hop_size: 480
win_size: 1920
fmin: 0
fmax: null
center: False
@@ -124,7 +124,7 @@ hifigan: !new:cosyvoice.hifigan.hifigan.HiFiGan
generator: !ref <hift>
discriminator: !new:cosyvoice.hifigan.discriminator.MultipleDiscriminator
mpd: !new:matcha.hifigan.models.MultiPeriodDiscriminator
mrd: !new:cosyvoice.hifigan.discriminator.MultiResolutionDiscriminator
mrd: !new:cosyvoice.hifigan.discriminator.MultiResSpecDiscriminator
mel_spec_transform: [
!ref <mel_spec_transform1>
]

2
examples/libritts/cosyvoice2/run.sh
View File

@@ -71,7 +71,7 @@ if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
fi
# train llm
export CUDA_VISIBLE_DEVICES="2,3,4,5,6,7"
export CUDA_VISIBLE_DEVICES="0,1,2,3"
num_gpus=$(echo $CUDA_VISIBLE_DEVICES | awk -F "," '{print NF}')
job_id=1986
dist_backend="nccl"