fix: Repair argument passing in both Dynamo paths

- Pass-through new TRT args in export
- Pass-through build failures arg in compile

Author: gs-olive

py/torch_tensorrt/dynamo/backend/__init__.py

@@ -45,6 +45,7 @@ def compile(
     min_block_size=MIN_BLOCK_SIZE,
     torch_executed_ops=[],
     torch_executed_modules=[],
+    pass_through_build_failures=PASS_THROUGH_BUILD_FAILURES,
     **kwargs,
 ):
     if debug:
@@ -86,6 +87,7 @@ def compile(
         workspace_size=workspace_size,
         min_block_size=min_block_size,
         torch_executed_ops=torch_executed_ops,
+        pass_through_build_failures=pass_through_build_failures,
         **kwargs,
     )
 

py/torch_tensorrt/dynamo/fx_ts_compat/lower.py

@@ -48,7 +48,11 @@ def compile(
     save_timing_cache=False,
     cuda_graph_batch_size=-1,
     is_aten=False,
-    use_experimental_fx_rt=False,
+    explicit_batch_dimension=True,
+    use_experimental_rt=False,
+    max_aux_streams=None,
+    version_compatible=False,
+    optimization_level=None,
     num_avg_timing_iters=1,
     torch_executed_ops=[],
     torch_executed_modules=[],
@@ -67,13 +71,17 @@ def compile(
         timing_cache_prefix: Timing cache file name for timing cache used by fx2trt.
         save_timing_cache: Update timing cache with current timing cache data if set to True.
         cuda_graph_batch_size: Cuda graph batch size, default to be -1.
-        use_experimental_fx_rt: Uses the next generation TRTModule which supports both Python and TorchScript based execution (including in C++).
+        explicit_batch_dimension: Whether to specify an explicit batch dimension to TRT
+        use_experimental_rt: Uses the next generation TRTModule which supports both Python and TorchScript based execution (including in C++).
+        max_aux_streams: max number of aux stream to use
+        version_compatible: enable version compatible feature
+        optimization_level: builder optimization level
     Returns:
         A torch.nn.Module lowered by TensorRT.
     """
-    if use_experimental_fx_rt and not explicit_batch_dimension:
+    if use_experimental_rt and not explicit_batch_dimension:
         raise ValueError(
-            "The experimental unifed runtime only supports explicit batch. Please make sure to set explicit_batch_dimension=True when use_experimental_fx_rt=True"
+            "The experimental unifed runtime only supports explicit batch. Please make sure to set explicit_batch_dimension=True when use_experimental_rt=True"
         )
 
     logger.warn(
@@ -122,7 +130,11 @@ def compile(
         save_timing_cache=save_timing_cache,
         cuda_graph_batch_size=cuda_graph_batch_size,
         is_aten=is_aten,
-        use_experimental_rt=use_experimental_fx_rt,
+        explicit_batch_dimension=explicit_batch_dimension,
+        use_experimental_rt=use_experimental_rt,
+        max_aux_streams=max_aux_streams,
+        version_compatible=version_compatible,
+        optimization_level=optimization_level,
     )
     lowerer = Lowerer.create(lower_setting=lower_setting)
     return lowerer(module, inputs)