deepgeek/CosyVoice
2
0
Fork 0
mirror of https://github.com/FunAudioLLM/CosyVoice.git synced 2026-02-04 17:39:25 +08:00
Code Issues Packages Projects Releases Wiki Activity

add flow cache inference code

Browse Source
This commit is contained in:
lyuxiang.lx
2025-04-07 21:23:09 +08:00
parent a69b7e275d
commit 39ffc50dec
4 changed files with 19 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
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: 2 # streaming inference chunk size, in second
num_decoding_left_chunks: 1 # streaming inference flow decoder left chunk size
chunk_size: 25 # streaming inference chunk size, in token
num_decoding_left_chunks: 1 # streaming inference flow decoder left chunk size, <0 means use all left chunks
# 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.
@@ -60,7 +60,7 @@ flow: !new:cosyvoice.flow.flow.CausalMaskedDiffWithXvec
input_size: 512
use_cnn_module: False
macaron_style: False
static_chunk_size: !ref <chunk_size> * <token_frame_rate>
static_chunk_size: !ref <chunk_size>
decoder: !new:cosyvoice.flow.flow_matching.CausalConditionalCFM
in_channels: 240
n_spks: 1
@@ -83,7 +83,7 @@ flow: !new:cosyvoice.flow.flow.CausalMaskedDiffWithXvec
num_mid_blocks: 12
num_heads: 8
act_fn: 'gelu'
static_chunk_size: !ref <chunk_size> * <token_frame_rate> * <token_mel_ratio> # here we use static_chunk_size because we want to fix kv cache size during inference
static_chunk_size: !ref <chunk_size> * <token_mel_ratio>
num_decoding_left_chunks: !ref <num_decoding_left_chunks>
hift: !new:cosyvoice.hifigan.generator.HiFTGenerator