Remove workspace parameter from generate_trt_engine.py (nvidia-holosc…

…an#532)

* Remove workspace parameter as it is no longer supported in tensorrt:24.08

Signed-off-by: Victor Chang <[email protected]>

* Fix lint error

Signed-off-by: Victor Chang <[email protected]>

---------

Signed-off-by: Victor Chang <[email protected]>

utilities/generate_trt_engine.py

@@ -51,22 +51,19 @@ def get_gpu_info():
     return gpu_info_list
 
 
-def convert_onnx(input_file, output_file, workspace_size, fp16_enabled):
+def convert_onnx(input_file, output_file, fp16_enabled):
     """
     Convert ONNX model to TensorRT engine using trtexec command.
 
     Args:
         input_file (str): Path to the input ONNX file.
         output_file (str): Path to the output engine file.
-        workspace_size (int): Workspace size for TensorRT engine (in MiB).
         fp16_enabled (bool): Flag indicating whether to enable FP16 mode.
 
     Returns:
         None
     """
-    trtexec_cmd = (
-        f"trtexec --onnx='{input_file}' --workspace={workspace_size} --saveEngine='{output_file}'"
-    )
+    trtexec_cmd = f"trtexec --onnx='{input_file}' --saveEngine='{output_file}'"
     if fp16_enabled:
         trtexec_cmd += " --fp16"
     status = os.system(trtexec_cmd)
@@ -80,7 +77,6 @@ def main():
     parser = argparse.ArgumentParser()
     parser.add_argument("--input", help="Input file path")
     parser.add_argument("--output", help="Output file path")
-    parser.add_argument("--workspace", type=int, default=2048, help="Workspace size in MiB")
     parser.add_argument("--fp16", default=False, action="store_true", help="Enable FP16 mode")
     parser.add_argument(
         "--force", default=False, action="store_true", help="Force overwrite existing output file"
@@ -103,10 +99,9 @@ def main():
             file=sys.stderr,
         )
         return
-    workspace_size = args.workspace
     fp16_enabled = args.fp16
 
-    result_code = convert_onnx(input_file, output_file, workspace_size, fp16_enabled)
+    result_code = convert_onnx(input_file, output_file, fp16_enabled)
     if result_code != 0:
         print("TensorRT engine generation failed.", file=sys.stderr)
         print("Exit code is ", result_code)