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

update doc

Browse Source
This commit is contained in:
jhqxxx
2026-05-24 19:26:47 +08:00
parent 925e5dd7c0
commit 754145246c
8 changed files with 291 additions and 101 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test_voxcpm.rs
+29 -32
View File
@@ -1,18 +1,16 @@
use std::time::Instant;
use aha::models::voxcpm_refact::generate::VoxCPMGenerateRefact;
use aha::params::chat::ChatCompletionParameters;
use aha::{
models::{
GenerateModel,
voxcpm::{generate::VoxCPMGenerate, tokenizer::SingleChineseTokenizer},
},
models::{GenerateModel, voxcpm::tokenizer::SingleChineseTokenizer},
utils::audio_utils::{extract_and_save_audio_from_response, save_wav_mono},
};
use anyhow::{Ok, Result};
#[test]
fn voxcpm_use_message_generate() -> Result<()> {
// RUST_BACKTRACE=1 cargo test -F cuda voxcpm_use_message_generate -r -- --nocapture
// RUST_BACKTRACE=1 cargo test -F cuda --test test_voxcpm voxcpm_use_message_generate -r -- --nocapture
let save_dir =
aha::utils::get_default_save_dir().ok_or(anyhow::anyhow!("Failed to get save dir"))?;
let model_path = format!("{}/OpenBMB/VoxCPM-0.5B/", save_dir);
@@ -42,7 +40,8 @@ fn voxcpm_use_message_generate() -> Result<()> {
"#;
let mes: ChatCompletionParameters = serde_json::from_str(message)?;
let i_start = Instant::now();
let mut voxcpm_generate = VoxCPMGenerate::init(&model_path, None, None)?;
// let mut voxcpm_generate = VoxCPMGenerate::init(&model_path, None, None)?;
let mut voxcpm_generate = VoxCPMGenerateRefact::init(&model_path, None, None)?;
let i_duration = i_start.elapsed();
println!("Time elapsed in load model is: {:?}", i_duration);
@@ -54,7 +53,6 @@ fn voxcpm_use_message_generate() -> Result<()> {
}
let i_duration = i_start.elapsed();
println!("Time elapsed in generate is: {:?}", i_duration);
// save_wav_mono(&generate, "voxcpm.wav", 16000)?;
Ok(())
}
@@ -66,35 +64,16 @@ fn voxcpm_generate() -> Result<()> {
let model_path = format!("{}/OpenBMB/VoxCPM-0.5B/", save_dir);
let i_start = Instant::now();
let mut voxcpm_generate = VoxCPMGenerate::init(&model_path, None, None)?;
let mut voxcpm_generate = VoxCPMGenerateRefact::init(&model_path, None, None)?;
let i_duration = i_start.elapsed();
println!("Time elapsed in load model is: {:?}", i_duration);
// let i_start = Instant::now();
// let generate = voxcpm_generate.generate_simple("太阳当空照，花儿对我笑，小鸟说早早早".to_string())?;
// let generate = voxcpm_generate.inference(
// "VoxCPM is an innovative end-to-end TTS model from ModelBest, designed to generate highly realistic speech.".to_string(),
// Some("啥子小师叔，打狗还要看主人，你再要继续，我，就是你的对手".to_string()),
// Some("file://./assets/audio/voice_01.wav".to_string()),
// // Some("一定被灰太狼给吃了，我已经为他准备好了花圈了".to_string()),
// // Some("file://./assets/audio/voice_05.wav".to_string()),
// 2,
// 100,
// 10,
// 2.0,
// // false,
// 6.0,
// )?;
// 创建prompt_cache
voxcpm_generate.build_prompt_cache(
"啥子小师叔，打狗还要看主人，你再要继续，我，就是你的对手".to_string(),
"file://./assets/audio/voice_01.wav".to_string(),
)?;
// 使用prompt_cache生成语音
let i_start = Instant::now();
let generate = voxcpm_generate.generate_use_prompt_cache(
"太阳当空照，花儿对我笑，小鸟说早早早".to_string(),
// let generate = voxcpm_generate.generate_simple("太阳当空照，花儿对我笑，小鸟说早早早".to_string())?;
let generate = voxcpm_generate.inference(
"VoxCPM is an innovative end-to-end TTS model from ModelBest, designed to generate highly realistic speech.".to_string(),
Some("天雷滚滚我好怕怕，劈得我浑身掉渣渣。突破天劫我笑哈哈，逆天改命我吹喇叭，滴答滴答滴滴答".to_string()),
Some("https://package-release.coderbox.cn/aiway/test/other/%E5%93%AA%E5%90%92.wav".to_string()),
2,
100,
10,
@@ -103,6 +82,23 @@ fn voxcpm_generate() -> Result<()> {
6.0,
)?;
// 创建prompt_cache
// voxcpm_generate.build_prompt_cache(
// "啥子小师叔，打狗还要看主人，你再要继续，我，就是你的对手".to_string(),
// "file://./assets/audio/voice_01.wav".to_string(),
// )?;
// // 使用prompt_cache生成语音
// let i_start = Instant::now();
// let generate = voxcpm_generate.generate_use_prompt_cache(
// "太阳当空照，花儿对我笑，小鸟说早早早".to_string(),
// 2,
// 100,
// 10,
// 2.0,
// false,
// 6.0,
// )?;
let i_duration = i_start.elapsed();
println!("Time elapsed in generate is: {:?}", i_duration);
save_wav_mono(&generate, "voxcpm.wav", 16000)?;
@@ -111,6 +107,7 @@ fn voxcpm_generate() -> Result<()> {
#[test]
fn voxcpm_tokenizer() -> Result<()> {
// RUST_BACKTRACE=1 cargo test -F cuda --test test_voxcpm voxcpm_tokenizer -r -- --nocapture
let save_dir =
aha::utils::get_default_save_dir().ok_or(anyhow::anyhow!("Failed to get save dir"))?;
let model_path = format!("{}/OpenBMB/VoxCPM-0.5B/", save_dir);
tests/test_voxcpm1_5.rs
+23 -30
View File
@@ -3,10 +3,7 @@ use std::time::Instant;
use aha::models::voxcpm_refact::generate::VoxCPMGenerateRefact;
use aha::params::chat::ChatCompletionParameters;
use aha::{
models::{
GenerateModel,
voxcpm::{generate::VoxCPMGenerate, tokenizer::SingleChineseTokenizer},
},
models::{GenerateModel, voxcpm::tokenizer::SingleChineseTokenizer},
utils::audio_utils::{extract_and_save_audio_from_response, save_wav_mono},
};
use anyhow::{Ok, Result};
@@ -43,7 +40,7 @@ fn voxcpm1_5_use_message_generate() -> Result<()> {
"#;
let mes: ChatCompletionParameters = serde_json::from_str(message)?;
let i_start = Instant::now();
let mut voxcpm_generate = VoxCPMGenerate::init(&model_path, None, None)?;
let mut voxcpm_generate = VoxCPMGenerateRefact::init(&model_path, None, None)?;
let i_duration = i_start.elapsed();
println!("Time elapsed in load model is: {:?}", i_duration);
@@ -66,43 +63,39 @@ fn voxcpm1_5_generate() -> Result<()> {
let model_path = format!("{}/OpenBMB/VoxCPM1.5/", save_dir);
let i_start = Instant::now();
let mut voxcpm_generate = VoxCPMGenerate::init(&model_path, None, None)?;
let mut voxcpm_generate = VoxCPMGenerateRefact::init(&model_path, None, None)?;
let i_duration = i_start.elapsed();
println!("Time elapsed in load model is: {:?}", i_duration);
let i_start = Instant::now();
// let generate = voxcpm_generate.generate_simple("太阳当空照，花儿对我笑，小鸟说早早早".to_string())?;
// let generate = voxcpm_generate.inference(
// "老大爷我来啦，红红火火恍恍惚惚".to_string(),
// Some("啥子小师叔，打狗还要看主人，你再要继续，我就是你的对手".to_string()),
// Some("file://./assets/audio/voice_01.wav".to_string()),
// // Some("一定被灰太狼给吃了，我已经为他准备好了花圈了".to_string()),
// // Some("file://./assets/audio/voice_05.wav".to_string()),
// 2,
// 4096,
// 10,
// 2.0,
// // false,
// 6.0,
// )?;
// 创建prompt_cache
voxcpm_generate.build_prompt_cache(
"啥子小师叔，打狗还要看主人，你再要继续，我，就是你的对手".to_string(),
"file://./assets/audio/voice_01.wav".to_string(),
)?;
// 使用prompt_cache生成语音
let generate = voxcpm_generate.generate_use_prompt_cache(
"太阳当空照，花儿对我笑，小鸟说早早早".to_string(),
let generate = voxcpm_generate.inference(
"老大爷我来啦，红红火火恍恍惚惚".to_string(),
Some("天雷滚滚我好怕怕，劈得我浑身掉渣渣。突破天劫我笑哈哈，逆天改命我吹喇叭，滴答滴答滴滴答".to_string()),
Some("https://package-release.coderbox.cn/aiway/test/other/%E5%93%AA%E5%90%92.wav".to_string()),
2,
100,
4096,
10,
2.0,
false,
6.0,
)?;
std::thread::sleep(std::time::Duration::from_secs(2));
// 创建prompt_cache
// voxcpm_generate.build_prompt_cache(
// "啥子小师叔，打狗还要看主人，你再要继续，我，就是你的对手".to_string(),
// "file://./assets/audio/voice_01.wav".to_string(),
// )?;
// // 使用prompt_cache生成语音
// let generate = voxcpm_generate.generate_use_prompt_cache(
// "太阳当空照，花儿对我笑，小鸟说早早早".to_string(),
// 2,
// 100,
// 10,
// 2.0,
// false,
// 6.0,
// )?;
let i_duration = i_start.elapsed();
println!("Time elapsed in generate is: {:?}", i_duration);
tests/test_voxcpm2.rs
+60 -9
View File
@@ -1,9 +1,12 @@
use std::time::Instant;
use aha::{
models::{GenerateModel, voxcpm::generate::VoxCPMGenerate},
models::{
GenerateModel, voxcpm::generate::VoxCPMGenerate,
voxcpm_refact::generate::VoxCPMGenerateRefact,
},
params::chat::ChatCompletionParameters,
utils::audio_utils::extract_and_save_audio_from_response,
utils::audio_utils::{extract_and_save_audio_from_response, save_wav_mono},
};
use anyhow::Result;
@@ -30,26 +33,74 @@ fn voxcpm2_use_message_generate() -> Result<()> {
},
{
"type": "text",
"text": "你好，这是aha在说话"
"text": "aha是一个基于Rust和Candle框架的本地AI推理引擎，支持多模态模型（文本、视觉、语音、OCR）。"
}
]
}
]
],
"metadata": {"prompt_text": "天雷滚滚我好怕怕，劈得我浑身掉渣渣。突破天劫我笑哈哈，逆天改命我吹喇叭，滴答滴答滴滴答"}
}
"#;
let mes: ChatCompletionParameters = serde_json::from_str(message)?;
let i_start = Instant::now();
let mut voxcpm_generate = VoxCPMGenerateRefact::init(&model_path, None, None)?;
let i_duration = i_start.elapsed();
println!("Time elapsed in load model is: {:?}", i_duration);
// for _ in 0..10 {
// let _ = voxcpm_generate.generate(mes.clone())?;
// }
// let mut times = vec![];
// for _ in 0..100 {
// let start = Instant::now();
// let _ = voxcpm_generate.generate(mes.clone())?;
// times.push(start.elapsed());
// }
// let mean = times.iter().sum::<Duration>() / 100;
// println!("mean: {:?}", mean);
// times.sort();
// println!("p99: {:?}", times[99]);
let i_start = Instant::now();
let generate = voxcpm_generate.generate(mes)?;
let i_duration = i_start.elapsed();
println!("Time elapsed in generate is: {:?}", i_duration);
let save_path = extract_and_save_audio_from_response(&generate, "./")?;
for path in save_path {
println!("save audio: {}", path);
}
Ok(())
}
#[test]
fn voxcpm2_generate() -> Result<()> {
// RUST_BACKTRACE=1 cargo test -F cuda --test test_voxcpm2 voxcpm2_generate -r -- --nocapture
let save_dir =
aha::utils::get_default_save_dir().ok_or(anyhow::anyhow!("Failed to get save dir"))?;
let model_path = format!("{}/OpenBMB/VoxCPM2/", save_dir);
let i_start = Instant::now();
let mut voxcpm_generate = VoxCPMGenerate::init(&model_path, None, None)?;
let i_duration = i_start.elapsed();
println!("Time elapsed in load model is: {:?}", i_duration);
let i_start = Instant::now();
let generate = voxcpm_generate.generate(mes)?;
let save_path = extract_and_save_audio_from_response(&generate, "./")?;
for path in save_path {
println!("save audio: {}", path);
}
let generate = voxcpm_generate.inference(
"aha是一个基于Rust和Candle框架的本地AI推理引擎，支持多模态模型（文本、视觉、语音、OCR）。"
.to_string(),
None,
None,
2,
1000,
10,
2.0,
6.0,
)?;
let i_duration = i_start.elapsed();
println!("Time elapsed in generate is: {:?}", i_duration);
save_wav_mono(
&generate,
"voxcpm2.wav",
voxcpm_generate.sample_rate() as u32,
)?;
Ok(())
}