update format

src/models/mod.rs

@@ -1,7 +1,9 @@
 pub mod qwen2_5vl;
 use anyhow::Result;
 use candle_core::{DType, Device};
-use openai_dive::v1::resources::chat::{ChatCompletionChunkResponse, ChatCompletionParameters, ChatCompletionResponse};
+use openai_dive::v1::resources::chat::{
+    ChatCompletionChunkResponse, ChatCompletionParameters, ChatCompletionResponse,
+};
 use rocket::futures::Stream;
 
 pub trait GenerateModel {
@@ -9,6 +11,8 @@ pub trait GenerateModel {
     where
         Self: Sized;
     fn generate(&mut self, mes: ChatCompletionParameters) -> Result<ChatCompletionResponse>;
-    fn generate_stream(&mut self, mes: ChatCompletionParameters) -> Result<impl Stream<Item = Result<ChatCompletionChunkResponse, anyhow::Error>>>;
+    fn generate_stream(
+        &mut self,
+        mes: ChatCompletionParameters,
+    ) -> Result<impl Stream<Item = Result<ChatCompletionChunkResponse, anyhow::Error>>>;
 }
-

src/models/qwen2_5vl/mod.rs

@@ -1,4 +1,4 @@
 pub mod config;
+pub mod generate;
 pub mod model;
 pub mod processor;
-pub mod generate;

src/tokenizer/mod.rs

@@ -1 +1 @@
-pub mod tokenizer;
+pub mod tokenizer;

src/utils/audio_utils.rs

@@ -0,0 +1 @@
+

src/utils/mod.rs

@@ -1,4 +1,4 @@
+pub mod img_utils;
 pub mod tensor_utils;
 pub mod utils;
-pub mod img_utils;
+pub mod video_utils;
-pub mod video_utils;

src/utils/utils.rs

@@ -3,7 +3,9 @@ use candle_core::{DType, Device};
 use candle_transformers::generation::LogitsProcessor;
 use openai_dive::v1::resources::{
     chat::{
-        ChatCompletionChoice, ChatCompletionChunkChoice, ChatCompletionChunkResponse, ChatCompletionResponse, ChatMessage, ChatMessageContent, DeltaChatMessage, DeltaFunction, DeltaToolCall, Function, ToolCall
+        ChatCompletionChoice, ChatCompletionChunkChoice, ChatCompletionChunkResponse,
+        ChatCompletionResponse, ChatMessage, ChatMessageContent, DeltaChatMessage, DeltaFunction,
+        DeltaToolCall, Function, ToolCall,
     },
     shared::FinishReason,
 };

tests/config_tests.rs

@@ -1,5 +1,5 @@
-use anyhow::Result;
 use aha::models::qwen2_5vl::config::Config;
+use anyhow::Result;
 
 #[test]
 fn qwen2_5vl_config() -> Result<()> {
@@ -9,4 +9,4 @@ fn qwen2_5vl_config() -> Result<()> {
     let config: Config = serde_json::from_slice(&std::fs::read(config_path)?)?;
     println!("{:?}", config);
     Ok(())
-}
+}

tests/test_qwen2_5vl.rs

@@ -1,12 +1,14 @@
 use std::{pin::pin, time::Instant};
 
-use aha::{models::{qwen2_5vl::generate::Qwen2_5VLGenerateModel, GenerateModel}, ModelType};
+use aha::{
-use anyhow::{Result};
+    ModelType,
+    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;
 
-
 #[test]
 fn qwen2_5vl_generate() -> Result<()> {
     // test with cpu :(太慢了, : RUST_BACKTRACE=1 cargo test qwen2_5vl_generate -- --nocapture
@@ -16,7 +18,7 @@ fn qwen2_5vl_generate() -> Result<()> {
     let dtype = DType::BF16;
 
     let model_path = "/home/jhq/huggingface_model/Qwen/Qwen2.5-VL-3B-Instruct/";
-    
+
     let message = r#"
     {
         "model": "qwen2.5vl",
@@ -40,7 +42,7 @@ fn qwen2_5vl_generate() -> Result<()> {
         ]
     }
     "#;
-    let mes:ChatCompletionParameters = serde_json::from_str(message)?;
+    let mes: ChatCompletionParameters = serde_json::from_str(message)?;
     let i_start = Instant::now();
     // let mut model = Qwen2_5VLGenerateModel::init(model_path, &device, dtype)?;
     let mut model = ModelType::init(ModelType::Qwen2_5VL, model_path, None, None)?;
@@ -49,10 +51,9 @@ fn qwen2_5vl_generate() -> Result<()> {
 
     let i_start = Instant::now();
     let result = model.generate(mes)?;
-    println!("generate: \n {:?}", result);    
+    println!("generate: \n {:?}", result);
     let i_duration = i_start.elapsed();
     println!("Time elapsed in generate is: {:?}", i_duration);
-    
 
     Ok(())
 }
@@ -64,7 +65,7 @@ async fn qwen2_5vl_stream() -> Result<()> {
     let dtype = DType::BF16;
 
     let model_path = "/home/jhq/huggingface_model/Qwen/Qwen2.5-VL-3B-Instruct/";
-    
+
     let message = r#"
     {
         "model": "qwen2.5vl",
@@ -88,7 +89,7 @@ async fn qwen2_5vl_stream() -> Result<()> {
         ]
     }
     "#;
-    let mes:ChatCompletionParameters = serde_json::from_str(message)?;
+    let mes: ChatCompletionParameters = serde_json::from_str(message)?;
     let i_start = Instant::now();
     // let mut model = Qwen2_5VLGenerateModel::init(model_path, &device, dtype)?;
     let mut model = ModelType::init(ModelType::Qwen2_5VL, model_path, None, None)?;
@@ -98,11 +99,11 @@ async fn qwen2_5vl_stream() -> Result<()> {
     let i_start = Instant::now();
     let mut stream = pin!(model.generate_stream(mes)?);
     while let Some(item) = stream.next().await {
-        println!("generate: \n {:?}", item);    
+        println!("generate: \n {:?}", item);
     }
-    
+
     let i_duration = i_start.elapsed();
     println!("Time elapsed in generate is: {:?}", i_duration);
-    
+
     Ok(())
-}
+}