add moss-tts-nano

tests/config_tests.rs

@@ -1,5 +1,5 @@
 use aha::models::{
-    deepseek_ocr::config::DeepseekOCRConfig, hunyuan_ocr::config::HunYuanVLConfig, lfm2::config::Lfm2Config, lfm2vl::config::{Lfm2ProcessorConfig, Lfm2VLConfig}, minicpm4::config::MiniCPM4Config, moss::config::{MossAudioTokenizerConfig, MossTTSConfig}, paddleocr_vl::config::PaddleOCRVLConfig, qwen2_5vl::config::Qwen2_5VLConfig, qwen3vl::config::Qwen3VLConfig, voxcpm::config::VoxCPMConfig
+    deepseek_ocr::config::DeepseekOCRConfig, hunyuan_ocr::config::HunYuanVLConfig, lfm2::config::Lfm2Config, lfm2vl::config::{Lfm2ProcessorConfig, Lfm2VLConfig}, minicpm4::config::MiniCPM4Config, moss_audio_tokenizer_nano::config::MossAudioTokenizerConfig, moss_tts_nano::config::MossTTSConfig, paddleocr_vl::config::PaddleOCRVLConfig, qwen2_5vl::config::Qwen2_5VLConfig, qwen3vl::config::Qwen3VLConfig, voxcpm::config::VoxCPMConfig
 };
 use anyhow::Result;
 
@@ -128,4 +128,4 @@ fn moss_tts_config() -> Result<()> {
     let config: MossTTSConfig = serde_json::from_slice(&std::fs::read(config_path)?)?;
     println!("{:?}", config);
     Ok(())
-}
+}

tests/test_moss_tts.rs

@@ -1,4 +1,4 @@
-use aha::models::moss::generate::MossTTSGenerate;
+use aha::models::moss_tts_nano::generate::MossTTSGenerate;
 use anyhow::Result;
 
 #[test]
@@ -10,11 +10,13 @@ fn moss_tts() -> Result<()> {
     let audio_tokenizer_path = format!("{}/openmoss/MOSS-Audio-Tokenizer-Nano/", save_dir);
     let mut model = MossTTSGenerate::init(&tts_path, &audio_tokenizer_path, None, None)?;
     let _ = model.generate(
-        "您好啊,吃饭了吗,吃的啥啊中午",
-        Some("file://./assets/audio/jiangjiang.wav"),
-        Some("哈喽大家好，我是蒋蒋"),
-        Some(aha::models::moss::tts_nano::MossTTSMode::Continuation),
+        "你在干吗啊",
         // None,
+        Some("file://./assets/audio/jiangjiang.wav"),
+        None,
+        // Some("哈喽大家好，我是蒋蒋"),
+        // Some(aha::models::moss_tts_nano::model::MossTTSMode::Continuation),
+        None,
     )?;
     Ok(())
 }

tests/test_voxcpm.rs

@@ -6,7 +6,7 @@ use aha::{
         GenerateModel,
         voxcpm::{generate::VoxCPMGenerate, tokenizer::SingleChineseTokenizer},
     },
-    utils::audio_utils::{extract_and_save_audio_from_response, save_wav},
+    utils::audio_utils::{extract_and_save_audio_from_response, save_wav_mono},
 };
 use anyhow::{Ok, Result};
 
@@ -54,7 +54,7 @@ fn voxcpm_use_message_generate() -> Result<()> {
     }
     let i_duration = i_start.elapsed();
     println!("Time elapsed in generate is: {:?}", i_duration);
-    // save_wav(&generate, "voxcpm.wav", 16000)?;
+    // save_wav_mono(&generate, "voxcpm.wav", 16000)?;
     Ok(())
 }
 
@@ -105,7 +105,7 @@ fn voxcpm_generate() -> Result<()> {
 
     let i_duration = i_start.elapsed();
     println!("Time elapsed in generate is: {:?}", i_duration);
-    save_wav(&generate, "voxcpm.wav", 16000)?;
+    save_wav_mono(&generate, "voxcpm.wav", 16000)?;
     Ok(())
 }
 

tests/test_voxcpm1_5.rs

@@ -7,7 +7,7 @@ use aha::{
         GenerateModel,
         voxcpm::{generate::VoxCPMGenerate, tokenizer::SingleChineseTokenizer},
     },
-    utils::audio_utils::{extract_and_save_audio_from_response, save_wav},
+    utils::audio_utils::{extract_and_save_audio_from_response, save_wav_mono},
 };
 use anyhow::{Ok, Result};
 
@@ -106,7 +106,7 @@ fn voxcpm1_5_generate() -> Result<()> {
 
     let i_duration = i_start.elapsed();
     println!("Time elapsed in generate is: {:?}", i_duration);
-    save_wav(&generate, "voxcpm1_5.wav", 44100)?;
+    save_wav_mono(&generate, "voxcpm1_5.wav", 44100)?;
     Ok(())
 }
 
@@ -169,7 +169,7 @@ fn voxcpm_refact_generate() -> Result<()> {
     std::thread::sleep(std::time::Duration::from_secs(2));
     let i_duration = i_start.elapsed();
     println!("Time elapsed in generate is: {:?}", i_duration);
-    save_wav(
+    save_wav_mono(
         &generate,
         "voxcpm.wav",
         voxcpm_generate.sample_rate() as u32,

tests/weight_test.rs

@@ -444,11 +444,14 @@ fn moss_audio_tokenizer_nano_weight() -> Result<()> {
     // cargo test -F cuda --test weight_test moss_audio_tokenizer_nano_weight -r -- --nocapture
     let save_dir =
         aha::utils::get_default_save_dir().ok_or(anyhow::anyhow!("Failed to get save dir"))?;
-    let model_path = format!("{}/openmoss/MOSS-Audio-Tokenizer-Nano/model-00001-of-00001.safetensors", save_dir);
+    let model_path = format!(
+        "{}/openmoss/MOSS-Audio-Tokenizer-Nano/model-00001-of-00001.safetensors",
+        save_dir
+    );
     let device = get_device(None);
     let weights = safetensors::load(model_path, &device)?;
     for (key, tensor) in weights.iter() {
         println!("=== {} === {:?}", key, tensor);
     }
     Ok(())
-}
+}