unify cargo version and add some ci rules

tests/config_tests.rs

@@ -1,4 +1,7 @@
-use aha::models::{minicpm4::config::MiniCPM4Config, qwen2_5vl::config::Qwen2_5VLConfig, voxcpm::config::VoxCPMConfig};
+use aha::models::{
+    minicpm4::config::MiniCPM4Config, qwen2_5vl::config::Qwen2_5VLConfig,
+    voxcpm::config::VoxCPMConfig,
+};
 use anyhow::Result;
 
 #[test]
@@ -11,7 +14,6 @@ fn qwen2_5_vl_config() -> Result<()> {
     Ok(())
 }
 
-
 #[test]
 fn minicpm4_config() -> Result<()> {
     // cargo test -F cuda,flash-attn minicpm4_config -- --nocapture
@@ -30,4 +32,4 @@ fn voxcpm_config() -> Result<()> {
     let config: VoxCPMConfig = serde_json::from_slice(&std::fs::read(config_path)?)?;
     println!("{:?}", config);
     Ok(())
-}
+}

tests/messy_test.rs

@@ -1,12 +1,12 @@
-use aha::utils::audio_utils::{load_audio_with_resample};
+use aha::utils::audio_utils::load_audio_with_resample;
 use anyhow::Result;
 
 #[test]
 fn messy_test() -> Result<()> {
     let device = candle_core::Device::Cpu;
     let wav_path = "./assets/audio/example.wav";
-    let audio_tensor = load_audio_with_resample(wav_path, device,Some(16000))?;
-    
+    let audio_tensor = load_audio_with_resample(wav_path, device, Some(16000))?;
+
     println!("audio_tensor: {}", audio_tensor);
     // let string = "你好啊".to_string();
     // let vec_str: Vec<String>= string.chars().map(|c| c.to_string()).collect();

tests/test_minicpm4.rs

@@ -1,8 +1,7 @@
 use std::{pin::pin, time::Instant};
 
-use aha::models::{minicpm4::generate::MiniCPMGenerateModel, GenerateModel};
+use aha::models::{GenerateModel, minicpm4::generate::MiniCPMGenerateModel};
 use anyhow::Result;
-use candle_core::{DType, Device};
 use openai_dive::v1::resources::chat::ChatCompletionParameters;
 use rocket::futures::StreamExt;
 
@@ -11,7 +10,7 @@ fn minicpm_generate() -> Result<()> {
     // test with cpu :(太慢了, : RUST_BACKTRACE=1 cargo test minicpm_generate -- --nocapture
     // test with cuda: RUST_BACKTRACE=1 cargo test -F cuda minicpm_generate -- --nocapture
     // test with cuda+flash-attn: RUST_BACKTRACE=1 cargo test -F cuda,flash-attn minicpm_generate -- --nocapture
-    
+
     let model_path = "/home/jhq/huggingface_model/OpenBMB/MiniCPM4-0.5B/";
     let message = r#"
     {
@@ -44,7 +43,7 @@ fn minicpm_generate() -> Result<()> {
 #[tokio::test]
 async fn minicpm_stream() -> Result<()> {
     // test with cuda+flash-attn: RUST_BACKTRACE=1 cargo test -F cuda,flash-attn minicpm_stream -- --nocapture
-    
+
     let model_path = "/home/jhq/huggingface_model/OpenBMB/MiniCPM4-0.5B/";
 
     let message = r#"

tests/test_qwen2_5vl.rs

@@ -1,10 +1,7 @@
 use std::{pin::pin, time::Instant};
 
-use aha::{
-    models::{GenerateModel, qwen2_5vl::generate::Qwen2_5VLGenerateModel},
-};
+use aha::models::{GenerateModel, qwen2_5vl::generate::Qwen2_5VLGenerateModel};
 use anyhow::Result;
-use candle_core::{DType, Device};
 use openai_dive::v1::resources::chat::ChatCompletionParameters;
 use rocket::futures::StreamExt;
 

tests/test_voxcpm.rs

@@ -1,20 +1,16 @@
-use anyhow::{Ok, Result};
-use std::{time::Instant};
+use std::time::Instant;
 
 use aha::{
-    models::voxcpm::{ generate::VoxCPMGenerate,
-        tokenizer::SingleChineseTokenizer,
-    },
-    utils::{
-        audio_utils::save_wav,
-    },
+    models::voxcpm::{generate::VoxCPMGenerate, tokenizer::SingleChineseTokenizer},
+    utils::audio_utils::save_wav,
 };
+use anyhow::{Ok, Result};
 
 #[test]
 fn voxcpm_generate() -> Result<()> {
     // RUST_BACKTRACE=1 cargo test -F cuda,flash-attn voxcpm_generate -- --nocapture
     let model_path = "/home/jhq/huggingface_model/openbmb/VoxCPM-0.5B/";
-    
+
     let i_start = Instant::now();
     let mut voxcpm_generate = VoxCPMGenerate::init(model_path, None, None)?;
     let i_duration = i_start.elapsed();
@@ -54,7 +50,7 @@ fn voxcpm_generate() -> Result<()> {
 
     let i_duration = i_start.elapsed();
     println!("Time elapsed in generate is: {:?}", i_duration);
-    let _ = save_wav(&generate, "voxcpm.wav")?;
+    save_wav(&generate, "voxcpm.wav")?;
     Ok(())
 }
 

tests/weight_test.rs

@@ -1,14 +1,14 @@
 use std::collections::HashMap;
 
-use aha::utils::utils::{find_type_files, get_device};
+use aha::utils::{find_type_files, get_device};
 use anyhow::Result;
-use candle_core::{pickle::{read_all_with_key, read_pth_tensor_info, PthTensors}, safetensors, Device, Tensor};
+use candle_core::{Device, pickle::read_all_with_key, safetensors};
 use candle_nn::VarBuilder;
 
 #[test]
 fn minicpm4_weight() -> Result<()> {
     let model_path = "/home/jhq/huggingface_model/OpenBMB/MiniCPM4-0.5B/";
-    let model_list = find_type_files(&model_path, "safetensors")?;
+    let model_list = find_type_files(model_path, "safetensors")?;
     let device = Device::Cpu;
     for m in model_list {
         let weights = safetensors::load(m, &device)?;
@@ -24,21 +24,24 @@ fn minicpm4_weight() -> Result<()> {
 #[test]
 fn voxcpm_weight() -> Result<()> {
     let model_path = "/home/jhq/huggingface_model/openbmb/VoxCPM-0.5B/";
-    let model_list = find_type_files(&model_path, "pth")?;
+    let model_list = find_type_files(model_path, "pth")?;
     println!("model_list: {:?}", model_list);
     let dev = get_device(None);
     let mut dict_to_hashmap = HashMap::new();
     let mut dtype = candle_core::DType::F16;
-    for m in model_list {  
-        let dict = read_all_with_key(m, Some("state_dict"))?;        
+    for m in model_list {
+        let dict = read_all_with_key(m, Some("state_dict"))?;
         dtype = dict[0].1.dtype();
         for (k, v) in dict {
             println!("key: {}, tensor shape: {:?}", k, v);
             dict_to_hashmap.insert(k, v);
-        }        
+        }
     }
     let vb = VarBuilder::from_tensors(dict_to_hashmap, dtype, &dev);
     let contain_key = vb.contains_tensor("encoder.block.4.block.2.block.3.weight_g");
-    println!("contain encoder.block.4.block.2.block.3.weight_g: {}", contain_key);
+    println!(
+        "contain encoder.block.4.block.2.block.3.weight_g: {}",
+        contain_key
+    );
     Ok(())
 }