diff --git a/alt_e2eshark/e2e_testing/backends.py b/alt_e2eshark/e2e_testing/backends.py
index 8d8daf44..bc226fb0 100644
--- a/alt_e2eshark/e2e_testing/backends.py
+++ b/alt_e2eshark/e2e_testing/backends.py
@@ -35,35 +35,13 @@ class SimpleIREEBackend(BackendBase):
     def __init__(self, *, device="local-task", hal_target_backend="llvm-cpu", extra_args : List[str] = None):
         self.device = device
         self.hal_target_backend = hal_target_backend
+        self.extra_args = []
         if extra_args:
-            self.extra_args = []
             for a in extra_args:
                 if a[0:2] == "--":
                     self.extra_args.append(a)
                 else:
                     self.extra_args.append("--" + a)
-        elif hal_target_backend == "rocm":
-            # some extra args for Mi300x - some of these may not work for other chips
-            self.extra_args = [
-                "--iree-rocm-target-chip=gfx942",
-                # "--iree-global-opt-propagate-transposes=true",
-                # "--iree-opt-outer-dim-concat=true",
-                # "--iree-opt-const-eval=false",
-                # "--iree-rocm-waves-per-eu=2",
-                # "--iree-llvmgpu-enable-prefetch",
-                # "--iree-flow-enable-aggressive-fusion",
-                # "--iree-flow-enable-fuse-horizontal-contractions=true",
-                # "--iree-opt-aggressively-propagate-transposes=true",
-                # "--iree-codegen-llvmgpu-use-vector-distribution=true",
-                # "--iree-preprocessing-pass-pipeline=builtin.module(util.func(iree-preprocessing-pad-to-intrinsics{pad-target-type=conv}))",
-                # maybe add iree-preprocessing-transpose-convolution-pipeline to preprocessing pipeline.
-            ]
-        elif hal_target_backend == "llvm-cpu":
-            self.extra_args = [
-                "--iree-input-demote-i64-to-i32",
-                # "--iree-llvmcpu-fail-on-large-vector=0",
-                # "--iree-llvmcpu-stack-allocation-limit=300000",
-            ]
 
     def compile(self, module, *, save_to: str = None):
         # compile to a vmfb for llvm-cpu
@@ -160,12 +138,6 @@ def __init__(self, *, device="local-task", hal_target_device="llvm-cpu", extra_a
             self.extra_args = [
                 "--iree-hip-target=gfx90a",
             ]
-        elif hal_target_device == "llvm-cpu":
-            self.extra_args = [
-                "--iree-input-demote-i64-to-i32",
-                # "--iree-llvmcpu-fail-on-large-vector=0",
-                # "--iree-llvmcpu-stack-allocation-limit=300000",
-            ]
         self.providers = ["IreeExecutionProvider"]
         # set provider options.
         provider_options_dict = dict()
diff --git a/alt_e2eshark/run.py b/alt_e2eshark/run.py
index bab1cd6d..f46c65fd 100644
--- a/alt_e2eshark/run.py
+++ b/alt_e2eshark/run.py
@@ -329,7 +329,7 @@ def _get_argparse():
         "--backend",
         choices=["llvm-cpu", "amd-aie", "rocm", "hip", "cuda", "vmvx", "metal-spirv", "vulkan-spirv"],
         default="llvm-cpu",
-        help="specifies the iree-hal-target-backend for compile phase",
+        help="specifies the iree-hal-target-device / iree-hal-target-backends for compile phase",
     )
     parser.add_argument(
         "-ica",