Update trt-ep setting (#229)

Author: jamjamjon

src/config/impl_ep.rs

@@ -122,8 +122,13 @@ impl Config {
     impl_ep_field!(tensorrt, max_workspace_size, usize);
     impl_ep_field!(tensorrt, min_subgraph_size, usize);
     impl_ep_field!(tensorrt, fp16, bool);
+    impl_ep_field!(tensorrt, int8, bool);
     impl_ep_field!(tensorrt, engine_cache, bool);
     impl_ep_field!(tensorrt, timing_cache, bool);
+    impl_ep_field!(tensorrt, dla, bool);
+    impl_ep_field!(tensorrt, dla_core, u32);
+    impl_ep_field!(tensorrt, int8_use_native_calibration_table, bool);
+    impl_ep_field!(tensorrt, detailed_build_log, bool);
 
     // ---------------- coreml -----------------------
     impl_ep_field!(coreml, static_input_shapes, bool);

src/ort/engine.rs

@@ -641,13 +641,20 @@ impl Engine {
                     let cache_path =
                         crate::Dir::Cache.crate_dir_default_with_subs(&["caches", "tensorrt"])?;
 
-                    let ep = ort::execution_providers::TensorRTExecutionProvider::default()
+                    let mut ep = ort::execution_providers::TensorRTExecutionProvider::default()
                         .with_device_id(id as i32)
                         .with_max_workspace_size(config.ep.tensorrt.max_workspace_size)
                         .with_builder_optimization_level(
                             config.ep.tensorrt.builder_optimization_level,
                         )
+                        .with_dla_core(config.ep.tensorrt.dla_core)
+                        .with_dla(config.ep.tensorrt.dla)
+                        .with_detailed_build_log(config.ep.tensorrt.detailed_build_log)
                         .with_fp16(config.ep.tensorrt.fp16)
+                        .with_int8(config.ep.tensorrt.int8)
+                        .with_int8_use_native_calibration_table(
+                            config.ep.tensorrt.int8_use_native_calibration_table,
+                        )
                         .with_engine_cache(config.ep.tensorrt.engine_cache)
                         .with_timing_cache(config.ep.tensorrt.timing_cache)
                         .with_dump_ep_context_model(config.ep.tensorrt.dump_ep_context_model)
@@ -659,6 +666,13 @@ impl Engine {
                         .with_profile_min_shapes(spec_min)
                         .with_profile_opt_shapes(spec_opt)
                         .with_profile_max_shapes(spec_max);
+                    if config.ep.tensorrt.int8
+                        && !config.ep.tensorrt.int8_calibration_table_name.is_empty()
+                    {
+                        ep = ep.with_int8_calibration_table_name(
+                            config.ep.tensorrt.int8_calibration_table_name.clone(),
+                        );
+                    }
 
                     match ep.is_available() {
                         Ok(true) => {

src/ort/ep_config.rs

@@ -45,27 +45,49 @@ impl Default for CudaConfig {
 /// NVIDIA TensorRT execution provider configuration.
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct TensorRtConfig {
+    pub detailed_build_log: bool,
     pub fp16: bool,
+    pub int8: bool,
+    /// Description: specify INT8 calibration table file for non-QDQ models in INT8 mode.
+    /// Note: calibration table should not be provided for QDQ model
+    /// because TensorRT doesn’t allow calibration table to be loded if there is any Q/DQ node in the model.
+    /// By default the name is empty.
+    pub int8_calibration_table_name: String,
+    /// Description: select what calibration table is used for non-QDQ models in INT8 mode.
+    /// If True, native TensorRT generated calibration table is used;
+    /// If False, ONNXRUNTIME tool generated calibration table is used.
+    /// Note: Please copy up-to-date calibration table file to trt_engine_cache_path before inference.
+    /// Calibration table is specific to models and calibration data sets.
+    /// Whenever new calibration table is generated, old file in the path should be cleaned up or be replaced.
+    pub int8_use_native_calibration_table: bool,
     pub engine_cache: bool,
     pub timing_cache: bool,
     pub dump_ep_context_model: bool,
     pub dump_subgraphs: bool,
     pub builder_optimization_level: u8,
     pub max_workspace_size: usize,
     pub min_subgraph_size: usize,
+    pub dla_core: u32,
+    pub dla: bool,
 }
 
 impl Default for TensorRtConfig {
     fn default() -> Self {
         Self {
+            detailed_build_log: true,
             fp16: true,
+            int8: false,
+            int8_calibration_table_name: "".to_string(),
+            int8_use_native_calibration_table: true,
             engine_cache: true,
             timing_cache: false,
             dump_ep_context_model: false, // TODO
             dump_subgraphs: false,
             builder_optimization_level: 3,  // 3, 0-5
             max_workspace_size: 1073741824, // 1G
             min_subgraph_size: 1,
+            dla_core: 0,
+            dla: false,
         }
     }
 }