tckm.yan/aha
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

add image api and audio api

Browse Source
This commit is contained in:
jhqxxx
2025-12-25 20:25:52 +08:00
parent bf53d3d869
commit 9a4fa636be
16 changed files with 694 additions and 72 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/models/voxcpm/model.rs
+6 -3
View File
@@ -513,7 +513,7 @@ impl VoxCPMModel {
let text_token = Tensor::cat(&[text_token, audio_start], D::Minus1)?;
let text_length = text_token.dim(0)?;
let mut audio =
load_audio_with_resample(path, self.device.clone(), Some(self.sample_rate))?;
load_audio_with_resample(&path, self.device.clone(), Some(self.sample_rate))?;
let patch_len = self.patch_size * self.chunk_size;
if audio.dim(1)? % patch_len != 0 {
audio = audio.pad_with_zeros(
@@ -728,8 +728,11 @@ impl VoxCPMModel {
) -> Result<HashMap<String, Tensor>> {
let text_token = self.tokenizer.encode(prompt_text)?;
let text_token = Tensor::from_slice(&text_token, text_token.len(), &self.device)?;
let mut audio =
load_audio_with_resample(prompt_wav_path, self.device.clone(), Some(self.sample_rate))?;
let mut audio = load_audio_with_resample(
&prompt_wav_path,
self.device.clone(),
Some(self.sample_rate),
)?;
let patch_len = self.patch_size * self.chunk_size;
if audio.dim(1)? % patch_len != 0 {
audio = audio.pad_with_zeros(D::Minus1, 0, patch_len - audio.dim(1)? % patch_len)?;