diff --git a/iree_tests/configs/config_onnx_cpu_llvm_sync.json b/iree_tests/configs/config_onnx_cpu_llvm_sync.json
index 3fd1ff69f..537472a69 100644
--- a/iree_tests/configs/config_onnx_cpu_llvm_sync.json
+++ b/iree_tests/configs/config_onnx_cpu_llvm_sync.json
@@ -12,8 +12,6 @@
   ],
   "skip_run_tests": [],
   "expected_compile_failures": [
-    "test_acosh",
-    "test_acosh_example",
     "test_adagrad",
     "test_adagrad_multiple",
     "test_adam",
@@ -31,10 +29,6 @@
     "test_ai_onnx_ml_label_encoder_string_int",
     "test_ai_onnx_ml_label_encoder_string_int_no_default",
     "test_ai_onnx_ml_label_encoder_tensor_value_only_mapping",
-    "test_asinh",
-    "test_asinh_example",
-    "test_atanh",
-    "test_atanh_example",
     "test_averagepool_2d_dilations",
     "test_averagepool_2d_pads",
     "test_averagepool_2d_precomputed_pads",
@@ -127,8 +121,6 @@
     "test_convtranspose_autopad_same",
     "test_convtranspose_kernel_shape",
     "test_convtranspose_output_shape",
-    "test_cosh",
-    "test_cosh_example",
     "test_cumsum_1d",
     "test_cumsum_1d_exclusive",
     "test_cumsum_1d_reverse",
@@ -141,6 +133,7 @@
     "test_dequantizelinear_axis",
     "test_dequantizelinear_blocked",
     "test_dequantizelinear_e4m3fn",
+    "test_dequantizelinear_e4m3fn_float16",
     "test_dequantizelinear_e4m3fn_zero_point",
     "test_dequantizelinear_e5m2",
     "test_dequantizelinear_int16",
@@ -161,15 +154,10 @@
     "test_equal_string",
     "test_equal_string_broadcast",
     "test_gathernd_example_int32_batch_dim1",
-    "test_globalmaxpool",
-    "test_globalmaxpool_precomputed",
     "test_gridsample_bicubic",
     "test_gridsample_bicubic_align_corners_0_additional_1",
     "test_gridsample_bicubic_align_corners_1_additional_1",
     "test_gridsample_border_padding",
-    "test_gridsample_nearest",
-    "test_gridsample_nearest_align_corners_0_additional_1",
-    "test_gridsample_nearest_align_corners_1_additional_1",
     "test_gridsample_reflection_padding",
     "test_gridsample_volumetric_bilinear_align_corners_0",
     "test_gridsample_volumetric_bilinear_align_corners_1",
@@ -215,18 +203,15 @@
     "test_lstm_defaults",
     "test_lstm_with_initial_bias",
     "test_lstm_with_peepholes",
-    "test_max_one_input",
     "test_maxpool_1d_default",
     "test_maxpool_2d_precomputed_same_upper",
     "test_maxpool_2d_same_lower",
     "test_maxpool_2d_same_upper",
     "test_maxpool_2d_uint8",
-    "test_maxpool_with_argmax_2d_precomputed_pads",
     "test_maxpool_with_argmax_2d_precomputed_strides",
     "test_maxunpool_export_with_output_shape",
     "test_maxunpool_export_without_output_shape",
     "test_melweightmatrix",
-    "test_min_one_input",
     "test_mod_mixed_sign_float16",
     "test_mod_mixed_sign_float32",
     "test_mod_mixed_sign_float64",
@@ -262,10 +247,8 @@
     "test_nonmaxsuppression_two_batches",
     "test_nonmaxsuppression_two_classes",
     "test_nonzero_example",
-    "test_onehot_negative_indices",
     "test_onehot_with_axis",
     "test_onehot_with_negative_axis",
-    "test_onehot_without_axis",
     "test_optional_get_element_tensor",
     "test_optional_has_element_empty_no_input_name_optional_input",
     "test_optional_has_element_empty_no_input_name_tensor_input",
@@ -318,9 +301,7 @@
     "test_reduce_log_sum_desc_axes_expanded",
     "test_reduce_log_sum_empty_set",
     "test_reduce_log_sum_empty_set_expanded",
-    "test_reduce_log_sum_exp_default_axes_keepdims_example",
     "test_reduce_log_sum_exp_default_axes_keepdims_example_expanded",
-    "test_reduce_log_sum_exp_default_axes_keepdims_random",
     "test_reduce_log_sum_exp_default_axes_keepdims_random_expanded",
     "test_reduce_log_sum_exp_do_not_keepdims_example",
     "test_reduce_log_sum_exp_do_not_keepdims_example_expanded",
@@ -393,15 +374,11 @@
     "test_resize_downsample_scales_cubic_A_n0p5_exclude_outside",
     "test_resize_downsample_scales_cubic_align_corners",
     "test_resize_downsample_scales_cubic_antialias",
-    "test_resize_downsample_scales_linear",
-    "test_resize_downsample_scales_linear_align_corners",
     "test_resize_downsample_scales_linear_antialias",
-    "test_resize_downsample_scales_linear_half_pixel_symmetric",
     "test_resize_downsample_scales_nearest",
     "test_resize_downsample_sizes_cubic",
     "test_resize_downsample_sizes_cubic_antialias",
     "test_resize_downsample_sizes_linear_antialias",
-    "test_resize_downsample_sizes_linear_pytorch_half_pixel",
     "test_resize_downsample_sizes_nearest",
     "test_resize_downsample_sizes_nearest_not_larger",
     "test_resize_downsample_sizes_nearest_not_smaller",
@@ -412,9 +389,6 @@
     "test_resize_upsample_scales_cubic_A_n0p5_exclude_outside",
     "test_resize_upsample_scales_cubic_align_corners",
     "test_resize_upsample_scales_cubic_asymmetric",
-    "test_resize_upsample_scales_linear",
-    "test_resize_upsample_scales_linear_align_corners",
-    "test_resize_upsample_scales_linear_half_pixel_symmetric",
     "test_resize_upsample_scales_nearest",
     "test_resize_upsample_scales_nearest_axes_2_3",
     "test_resize_upsample_scales_nearest_axes_3_2",
@@ -423,7 +397,6 @@
     "test_resize_upsample_sizes_nearest_axes_2_3",
     "test_resize_upsample_sizes_nearest_axes_3_2",
     "test_resize_upsample_sizes_nearest_ceil_half_pixel",
-    "test_resize_upsample_sizes_nearest_floor_align_corners",
     "test_resize_upsample_sizes_nearest_not_larger",
     "test_resize_upsample_sizes_nearest_round_prefer_ceil_asymmetric",
     "test_reversesequence_batch",
@@ -507,8 +480,6 @@
     "test_simple_rnn_batchwise",
     "test_simple_rnn_defaults",
     "test_simple_rnn_with_initial_bias",
-    "test_sinh",
-    "test_sinh_example",
     "test_slice",
     "test_slice_default_steps",
     "test_slice_end_out_of_bounds",
@@ -563,15 +534,6 @@
     "test_training_dropout_mask",
     "test_training_dropout_zero_ratio",
     "test_training_dropout_zero_ratio_mask",
-    "test_triu",
-    "test_triu_neg",
-    "test_triu_one_row",
-    "test_triu_out_neg_out",
-    "test_triu_out_pos",
-    "test_triu_pos",
-    "test_triu_square",
-    "test_triu_square_neg",
-    "test_triu_zero",
     "test_unique_not_sorted_without_axis",
     "test_unique_sorted_with_axis",
     "test_unique_sorted_with_axis_3d",
@@ -615,11 +577,15 @@
     "test_elu_default",
     "test_eyelike_with_dtype",
     "test_gather_elements_negative_indices",
+    "test_gridsample_nearest",
+    "test_gridsample_nearest_align_corners_0_additional_1",
+    "test_gridsample_nearest_align_corners_1_additional_1",
     "test_hardsigmoid",
     "test_hardsigmoid_default",
     "test_hardsigmoid_example",
     "test_hardswish_expanded",
     "test_max_float64",
+    "test_maxpool_2d_ceil_output_size_reduce_by_one",
     "test_min_float64",
     "test_mod_mixed_sign_int16",
     "test_mod_mixed_sign_int32",
@@ -638,8 +604,16 @@
     "test_qlinearmatmul_3D_int8_float32",
     "test_qlinearmatmul_3D_uint8_float16",
     "test_qlinearmatmul_3D_uint8_float32",
+    "test_reduce_log_sum_exp_default_axes_keepdims_example",
+    "test_reduce_log_sum_exp_default_axes_keepdims_random",
     "test_reduce_min_empty_set",
     "test_reduce_sum_empty_set_non_reduced_axis_zero",
+    "test_resize_downsample_scales_linear",
+    "test_resize_downsample_scales_linear_half_pixel_symmetric",
+    "test_resize_downsample_sizes_linear_pytorch_half_pixel",
+    "test_resize_upsample_scales_linear",
+    "test_resize_upsample_scales_linear_half_pixel_symmetric",
+    "test_resize_upsample_sizes_nearest_floor_align_corners",
     "test_scatter_elements_with_negative_indices",
     "test_sce_mean_no_weight_ii",
     "test_sce_mean_no_weight_ii_log_prob",
@@ -652,6 +626,7 @@
     "test_size_example",
     "test_split_zero_size_splits_opset13",
     "test_split_zero_size_splits_opset18",
-    "test_tril_zero"
+    "test_tril_zero",
+    "test_triu_zero"
   ]
 }
diff --git a/iree_tests/configs/config_pytorch_models_cpu_llvm_task.json b/iree_tests/configs/config_pytorch_models_cpu_llvm_task.json
index acc861820..c0f21b08a 100644
--- a/iree_tests/configs/config_pytorch_models_cpu_llvm_task.json
+++ b/iree_tests/configs/config_pytorch_models_cpu_llvm_task.json
@@ -15,5 +15,7 @@
       "opt-125M", // TODO(#17344): need to regenerate .mlirbc
       "resnet50", // TODO(#17344): need to regenerate .mlirbc
     ],
-    "expected_run_failures": []
+    "expected_run_failures": [
+      "sdxl-vae-decode-tank",
+    ]
   }
diff --git a/iree_tests/configs/config_sdxl_scheduled_unet_cpu_llvm_task.json b/iree_tests/configs/config_sdxl_scheduled_unet_cpu_llvm_task.json
index 31ffd338a..ec69c48ce 100644
--- a/iree_tests/configs/config_sdxl_scheduled_unet_cpu_llvm_task.json
+++ b/iree_tests/configs/config_sdxl_scheduled_unet_cpu_llvm_task.json
@@ -18,5 +18,7 @@
     "skip_compile_tests": [],
     "skip_run_tests": [],
     "expected_compile_failures": [],
-    "expected_run_failures": []
+    "expected_run_failures": [
+      "sdxl-scheduled-unet-3-tank",
+    ]
 }
diff --git a/iree_tests/onnx/node/generated/test_dequantizelinear_e4m3fn_float16/input_0.npy b/iree_tests/onnx/node/generated/test_dequantizelinear_e4m3fn_float16/input_0.npy
new file mode 100644
index 000000000..5b5fb425b
Binary files /dev/null and b/iree_tests/onnx/node/generated/test_dequantizelinear_e4m3fn_float16/input_0.npy differ
diff --git a/iree_tests/onnx/node/generated/test_dequantizelinear_e4m3fn_float16/input_1.npy b/iree_tests/onnx/node/generated/test_dequantizelinear_e4m3fn_float16/input_1.npy
new file mode 100644
index 000000000..b2f097179
Binary files /dev/null and b/iree_tests/onnx/node/generated/test_dequantizelinear_e4m3fn_float16/input_1.npy differ
diff --git a/iree_tests/onnx/node/generated/test_dequantizelinear_e4m3fn_float16/model.mlir b/iree_tests/onnx/node/generated/test_dequantizelinear_e4m3fn_float16/model.mlir
new file mode 100644
index 000000000..b374e7e54
--- /dev/null
+++ b/iree_tests/onnx/node/generated/test_dequantizelinear_e4m3fn_float16/model.mlir
@@ -0,0 +1,8 @@
+module {
+  func.func @test_dequantizelinear_e4m3fn_float16(%arg0: !torch.vtensor<[5],f8E4M3FN>, %arg1: !torch.vtensor<[],f16>) -> !torch.vtensor<[5],f16> attributes {torch.onnx_meta.ir_version = 10 : si64, torch.onnx_meta.opset_version = 21 : si64, torch.onnx_meta.producer_name = "backend-test", torch.onnx_meta.producer_version = ""} {
+    %none = torch.constant.none
+    %0 = torch.operator "onnx.DequantizeLinear"(%arg0, %arg1) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[5],f8E4M3FN>, !torch.vtensor<[],f16>) -> !torch.vtensor<[5],f16> 
+    return %0 : !torch.vtensor<[5],f16>
+  }
+}
+
diff --git a/iree_tests/onnx/node/generated/test_dequantizelinear_e4m3fn_float16/output_0.npy b/iree_tests/onnx/node/generated/test_dequantizelinear_e4m3fn_float16/output_0.npy
new file mode 100644
index 000000000..3b4489613
Binary files /dev/null and b/iree_tests/onnx/node/generated/test_dequantizelinear_e4m3fn_float16/output_0.npy differ
diff --git a/iree_tests/onnx/node/generated/test_dequantizelinear_e4m3fn_float16/test_data_flags.txt b/iree_tests/onnx/node/generated/test_dequantizelinear_e4m3fn_float16/test_data_flags.txt
new file mode 100644
index 000000000..37bc3f8a2
--- /dev/null
+++ b/iree_tests/onnx/node/generated/test_dequantizelinear_e4m3fn_float16/test_data_flags.txt
@@ -0,0 +1,3 @@
+--input=5xf32=@input_0.bin
+--input=xf16=@input_1.bin
+--expected_output=5xf16=@output_0.bin
diff --git a/iree_tests/onnx/node/generated/test_group_normalization_epsilon_expanded/model.mlir b/iree_tests/onnx/node/generated/test_group_normalization_epsilon_expanded/model.mlir
index 2e788b874..5d68211f9 100644
--- a/iree_tests/onnx/node/generated/test_group_normalization_epsilon_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_group_normalization_epsilon_expanded/model.mlir
@@ -24,17 +24,17 @@ module {
     %20 = torch.operator "onnx.Sqrt"(%19) : (!torch.vtensor<[?,?,1],f32>) -> !torch.vtensor<[?,?,1],f32> 
     %21 = torch.operator "onnx.Sub"(%12, %14) : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[?,?,1],f32>) -> !torch.vtensor<[?,?,?],f32> 
     %22 = torch.operator "onnx.Div"(%21, %20) : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[?,?,1],f32>) -> !torch.vtensor<[?,?,?],f32> 
-    %23 = torch.operator "onnx.Reshape"(%22, %3) : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[?,?,?,?],f32> 
-    %24 = torch.operator "onnx.Reshape"(%23, %11) : (!torch.vtensor<[?,?,?,?],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[?,?,?],f32> 
-    %25 = torch.operator "onnx.Cast"(%24) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[?,?,?],f32>) -> !torch.vtensor<[?,?,?],f32> 
+    %23 = torch.operator "onnx.Reshape"(%22, %3) : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[3,4,2,2],f32> 
+    %24 = torch.operator "onnx.Reshape"(%23, %11) : (!torch.vtensor<[3,4,2,2],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3,4,4],f32> 
+    %25 = torch.operator "onnx.Cast"(%24) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[3,4,4],f32>) -> !torch.vtensor<[3,4,4],f32> 
     %26 = torch.operator "onnx.Constant"() {torch.onnx.value_ints = [1 : si64, -1 : si64, 1 : si64]} : () -> !torch.vtensor<[3],si64> 
     %27 = torch.operator "onnx.Cast"(%arg1) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[4],f32>) -> !torch.vtensor<[4],f32> 
     %28 = torch.operator "onnx.Cast"(%arg2) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[4],f32>) -> !torch.vtensor<[4],f32> 
     %29 = torch.operator "onnx.Reshape"(%27, %26) : (!torch.vtensor<[4],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[1,4,1],f32> 
     %30 = torch.operator "onnx.Reshape"(%28, %26) : (!torch.vtensor<[4],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[1,4,1],f32> 
-    %31 = torch.operator "onnx.Mul"(%29, %25) : (!torch.vtensor<[1,4,1],f32>, !torch.vtensor<[?,?,?],f32>) -> !torch.vtensor<[?,4,?],f32> 
-    %32 = torch.operator "onnx.Add"(%31, %30) : (!torch.vtensor<[?,4,?],f32>, !torch.vtensor<[1,4,1],f32>) -> !torch.vtensor<[?,4,?],f32> 
-    %33 = torch.operator "onnx.Reshape"(%32, %3) : (!torch.vtensor<[?,4,?],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[3,4,2,2],f32> 
+    %31 = torch.operator "onnx.Mul"(%29, %25) : (!torch.vtensor<[1,4,1],f32>, !torch.vtensor<[3,4,4],f32>) -> !torch.vtensor<[3,4,4],f32> 
+    %32 = torch.operator "onnx.Add"(%31, %30) : (!torch.vtensor<[3,4,4],f32>, !torch.vtensor<[1,4,1],f32>) -> !torch.vtensor<[3,4,4],f32> 
+    %33 = torch.operator "onnx.Reshape"(%32, %3) : (!torch.vtensor<[3,4,4],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[3,4,2,2],f32> 
     return %33 : !torch.vtensor<[3,4,2,2],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_group_normalization_example_expanded/model.mlir b/iree_tests/onnx/node/generated/test_group_normalization_example_expanded/model.mlir
index 1718ecab6..e560e6162 100644
--- a/iree_tests/onnx/node/generated/test_group_normalization_example_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_group_normalization_example_expanded/model.mlir
@@ -24,17 +24,17 @@ module {
     %20 = torch.operator "onnx.Sqrt"(%19) : (!torch.vtensor<[?,?,1],f32>) -> !torch.vtensor<[?,?,1],f32> 
     %21 = torch.operator "onnx.Sub"(%12, %14) : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[?,?,1],f32>) -> !torch.vtensor<[?,?,?],f32> 
     %22 = torch.operator "onnx.Div"(%21, %20) : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[?,?,1],f32>) -> !torch.vtensor<[?,?,?],f32> 
-    %23 = torch.operator "onnx.Reshape"(%22, %3) : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[?,?,?,?],f32> 
-    %24 = torch.operator "onnx.Reshape"(%23, %11) : (!torch.vtensor<[?,?,?,?],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[?,?,?],f32> 
-    %25 = torch.operator "onnx.Cast"(%24) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[?,?,?],f32>) -> !torch.vtensor<[?,?,?],f32> 
+    %23 = torch.operator "onnx.Reshape"(%22, %3) : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[3,4,2,2],f32> 
+    %24 = torch.operator "onnx.Reshape"(%23, %11) : (!torch.vtensor<[3,4,2,2],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3,4,4],f32> 
+    %25 = torch.operator "onnx.Cast"(%24) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[3,4,4],f32>) -> !torch.vtensor<[3,4,4],f32> 
     %26 = torch.operator "onnx.Constant"() {torch.onnx.value_ints = [1 : si64, -1 : si64, 1 : si64]} : () -> !torch.vtensor<[3],si64> 
     %27 = torch.operator "onnx.Cast"(%arg1) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[4],f32>) -> !torch.vtensor<[4],f32> 
     %28 = torch.operator "onnx.Cast"(%arg2) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[4],f32>) -> !torch.vtensor<[4],f32> 
     %29 = torch.operator "onnx.Reshape"(%27, %26) : (!torch.vtensor<[4],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[1,4,1],f32> 
     %30 = torch.operator "onnx.Reshape"(%28, %26) : (!torch.vtensor<[4],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[1,4,1],f32> 
-    %31 = torch.operator "onnx.Mul"(%29, %25) : (!torch.vtensor<[1,4,1],f32>, !torch.vtensor<[?,?,?],f32>) -> !torch.vtensor<[?,4,?],f32> 
-    %32 = torch.operator "onnx.Add"(%31, %30) : (!torch.vtensor<[?,4,?],f32>, !torch.vtensor<[1,4,1],f32>) -> !torch.vtensor<[?,4,?],f32> 
-    %33 = torch.operator "onnx.Reshape"(%32, %3) : (!torch.vtensor<[?,4,?],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[3,4,2,2],f32> 
+    %31 = torch.operator "onnx.Mul"(%29, %25) : (!torch.vtensor<[1,4,1],f32>, !torch.vtensor<[3,4,4],f32>) -> !torch.vtensor<[3,4,4],f32> 
+    %32 = torch.operator "onnx.Add"(%31, %30) : (!torch.vtensor<[3,4,4],f32>, !torch.vtensor<[1,4,1],f32>) -> !torch.vtensor<[3,4,4],f32> 
+    %33 = torch.operator "onnx.Reshape"(%32, %3) : (!torch.vtensor<[3,4,4],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[3,4,2,2],f32> 
     return %33 : !torch.vtensor<[3,4,2,2],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis0_expanded/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis0_expanded/model.mlir
index d4d09357b..d6647792d 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis0_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis0_expanded/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<0> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[2],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[0],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[0],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[2],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[0],si64>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[2],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[3,4],f32>) -> !torch.vtensor<[1,12],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[1,12],f32>) -> !torch.vtensor<[1,12],f32> 
     %12 = torch.operator "onnx.ReduceMean"(%11) {torch.onnx.axes = [1 : si64]} : (!torch.vtensor<[1,12],f32>) -> !torch.vtensor<[1,1],f32> 
@@ -29,8 +29,8 @@ module {
     %25 = torch.operator "onnx.Add"(%23, %24) : (!torch.vtensor<[1,12],f32>, !torch.vtensor<[1,12],f32>) -> !torch.vtensor<[1,12],f32> 
     %26 = torch.operator "onnx.Reshape"(%25, %2) : (!torch.vtensor<[1,12],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,4],f32> 
     %27 = torch.operator "onnx.Reciprocal"(%18) : (!torch.vtensor<[1,1],f32>) -> !torch.vtensor<[1,1],f32> 
-    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[1,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[1,1],f32> 
+    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[1,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[1,1],f32> 
     return %26, %28, %29 : !torch.vtensor<[3,4],f32>, !torch.vtensor<[1,1],f32>, !torch.vtensor<[1,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis0_expanded_ver18/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis0_expanded_ver18/model.mlir
index 5123153e9..c841817b3 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis0_expanded_ver18/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis0_expanded_ver18/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<0> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[2],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[0],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[0],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[2],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[0],si64>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[2],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[3,4],f32>) -> !torch.vtensor<[1,12],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[1,12],f32>) -> !torch.vtensor<[1,12],f32> 
     %12 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<1> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
@@ -30,8 +30,8 @@ module {
     %26 = torch.operator "onnx.Add"(%24, %25) : (!torch.vtensor<[1,12],f32>, !torch.vtensor<[1,12],f32>) -> !torch.vtensor<[1,12],f32> 
     %27 = torch.operator "onnx.Reshape"(%26, %2) : (!torch.vtensor<[1,12],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,4],f32> 
     %28 = torch.operator "onnx.Reciprocal"(%19) : (!torch.vtensor<[1,1],f32>) -> !torch.vtensor<[1,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[1,1],f32> 
-    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[1,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[1,1],f32> 
+    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[1,1],f32> 
     return %27, %29, %30 : !torch.vtensor<[3,4],f32>, !torch.vtensor<[1,1],f32>, !torch.vtensor<[1,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis1_expanded/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis1_expanded/model.mlir
index 76134be00..85a4bd7a3 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis1_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis1_expanded/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<1> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[2],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[1],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[2],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 1 : si64} : (!torch.vtensor<[3,4],f32>) -> !torch.vtensor<[3,4],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[3,4],f32>) -> !torch.vtensor<[3,4],f32> 
     %12 = torch.operator "onnx.ReduceMean"(%11) {torch.onnx.axes = [1 : si64]} : (!torch.vtensor<[3,4],f32>) -> !torch.vtensor<[3,1],f32> 
@@ -29,8 +29,8 @@ module {
     %25 = torch.operator "onnx.Add"(%23, %24) : (!torch.vtensor<[3,4],f32>, !torch.vtensor<[1,4],f32>) -> !torch.vtensor<[3,4],f32> 
     %26 = torch.operator "onnx.Reshape"(%25, %2) : (!torch.vtensor<[3,4],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,4],f32> 
     %27 = torch.operator "onnx.Reciprocal"(%18) : (!torch.vtensor<[3,1],f32>) -> !torch.vtensor<[3,1],f32> 
-    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[3,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[3,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[3,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[3,1],f32> 
+    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[3,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[3,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,1],f32> 
     return %26, %28, %29 : !torch.vtensor<[3,4],f32>, !torch.vtensor<[3,1],f32>, !torch.vtensor<[3,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis1_expanded_ver18/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis1_expanded_ver18/model.mlir
index 2503c2c5e..593f397ab 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis1_expanded_ver18/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_2d_axis1_expanded_ver18/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<1> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[2],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[1],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[2],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 1 : si64} : (!torch.vtensor<[3,4],f32>) -> !torch.vtensor<[3,4],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[3,4],f32>) -> !torch.vtensor<[3,4],f32> 
     %12 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<1> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
@@ -30,8 +30,8 @@ module {
     %26 = torch.operator "onnx.Add"(%24, %25) : (!torch.vtensor<[3,4],f32>, !torch.vtensor<[1,4],f32>) -> !torch.vtensor<[3,4],f32> 
     %27 = torch.operator "onnx.Reshape"(%26, %2) : (!torch.vtensor<[3,4],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,4],f32> 
     %28 = torch.operator "onnx.Reciprocal"(%19) : (!torch.vtensor<[3,1],f32>) -> !torch.vtensor<[3,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[3,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[3,1],f32> 
-    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[3,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[3,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[3,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,1],f32> 
+    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[3,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,1],f32> 
     return %27, %29, %30 : !torch.vtensor<[3,4],f32>, !torch.vtensor<[3,1],f32>, !torch.vtensor<[3,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis0_epsilon_expanded/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis0_epsilon_expanded/model.mlir
index f2316cd6d..16d3f9928 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis0_epsilon_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis0_epsilon_expanded/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<0> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[3],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[0],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[0],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[3],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[0],si64>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[2,3,5],f32>) -> !torch.vtensor<[1,30],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[1,30],f32>) -> !torch.vtensor<[1,30],f32> 
     %12 = torch.operator "onnx.ReduceMean"(%11) {torch.onnx.axes = [1 : si64]} : (!torch.vtensor<[1,30],f32>) -> !torch.vtensor<[1,1],f32> 
@@ -29,8 +29,8 @@ module {
     %25 = torch.operator "onnx.Add"(%23, %24) : (!torch.vtensor<[1,30],f32>, !torch.vtensor<[1,30],f32>) -> !torch.vtensor<[1,30],f32> 
     %26 = torch.operator "onnx.Reshape"(%25, %2) : (!torch.vtensor<[1,30],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[2,3,5],f32> 
     %27 = torch.operator "onnx.Reciprocal"(%18) : (!torch.vtensor<[1,1],f32>) -> !torch.vtensor<[1,1],f32> 
-    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[1,1,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[1,1,1],f32> 
+    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[1,1,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[1,1,1],f32> 
     return %26, %28, %29 : !torch.vtensor<[2,3,5],f32>, !torch.vtensor<[1,1,1],f32>, !torch.vtensor<[1,1,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis0_epsilon_expanded_ver18/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis0_epsilon_expanded_ver18/model.mlir
index fc8c92eed..42d897c36 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis0_epsilon_expanded_ver18/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis0_epsilon_expanded_ver18/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<0> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[3],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[0],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[0],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[3],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[0],si64>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[2,3,5],f32>) -> !torch.vtensor<[1,30],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[1,30],f32>) -> !torch.vtensor<[1,30],f32> 
     %12 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<1> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
@@ -30,8 +30,8 @@ module {
     %26 = torch.operator "onnx.Add"(%24, %25) : (!torch.vtensor<[1,30],f32>, !torch.vtensor<[1,30],f32>) -> !torch.vtensor<[1,30],f32> 
     %27 = torch.operator "onnx.Reshape"(%26, %2) : (!torch.vtensor<[1,30],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[2,3,5],f32> 
     %28 = torch.operator "onnx.Reciprocal"(%19) : (!torch.vtensor<[1,1],f32>) -> !torch.vtensor<[1,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[1,1,1],f32> 
-    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[1,1,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[1,1,1],f32> 
+    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[1,1,1],f32> 
     return %27, %29, %30 : !torch.vtensor<[2,3,5],f32>, !torch.vtensor<[1,1,1],f32>, !torch.vtensor<[1,1,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis1_epsilon_expanded/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis1_epsilon_expanded/model.mlir
index 0513dd123..9fcf2c79c 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis1_epsilon_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis1_epsilon_expanded/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<1> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[3],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[1],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[2],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[1],si64>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 1 : si64} : (!torch.vtensor<[2,3,5],f32>) -> !torch.vtensor<[2,15],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[2,15],f32>) -> !torch.vtensor<[2,15],f32> 
     %12 = torch.operator "onnx.ReduceMean"(%11) {torch.onnx.axes = [1 : si64]} : (!torch.vtensor<[2,15],f32>) -> !torch.vtensor<[2,1],f32> 
@@ -29,8 +29,8 @@ module {
     %25 = torch.operator "onnx.Add"(%23, %24) : (!torch.vtensor<[2,15],f32>, !torch.vtensor<[1,15],f32>) -> !torch.vtensor<[2,15],f32> 
     %26 = torch.operator "onnx.Reshape"(%25, %2) : (!torch.vtensor<[2,15],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[2,3,5],f32> 
     %27 = torch.operator "onnx.Reciprocal"(%18) : (!torch.vtensor<[2,1],f32>) -> !torch.vtensor<[2,1],f32> 
-    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,1,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,1,1],f32> 
+    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[2,1,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[2,1,1],f32> 
     return %26, %28, %29 : !torch.vtensor<[2,3,5],f32>, !torch.vtensor<[2,1,1],f32>, !torch.vtensor<[2,1,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis1_epsilon_expanded_ver18/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis1_epsilon_expanded_ver18/model.mlir
index 35afd9b74..2e598232d 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis1_epsilon_expanded_ver18/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis1_epsilon_expanded_ver18/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<1> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[3],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[1],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[2],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[1],si64>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 1 : si64} : (!torch.vtensor<[2,3,5],f32>) -> !torch.vtensor<[2,15],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[2,15],f32>) -> !torch.vtensor<[2,15],f32> 
     %12 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<1> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
@@ -30,8 +30,8 @@ module {
     %26 = torch.operator "onnx.Add"(%24, %25) : (!torch.vtensor<[2,15],f32>, !torch.vtensor<[1,15],f32>) -> !torch.vtensor<[2,15],f32> 
     %27 = torch.operator "onnx.Reshape"(%26, %2) : (!torch.vtensor<[2,15],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[2,3,5],f32> 
     %28 = torch.operator "onnx.Reciprocal"(%19) : (!torch.vtensor<[2,1],f32>) -> !torch.vtensor<[2,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,1,1],f32> 
-    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,1,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[2,1,1],f32> 
+    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[2,1,1],f32> 
     return %27, %29, %30 : !torch.vtensor<[2,3,5],f32>, !torch.vtensor<[2,1,1],f32>, !torch.vtensor<[2,1,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis2_epsilon_expanded/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis2_epsilon_expanded/model.mlir
index 96ea9a1c9..6c74fb6d2 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis2_epsilon_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis2_epsilon_expanded/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<2> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[3],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[2],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[2],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[2],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[3],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[2,3,5],f32>) -> !torch.vtensor<[6,5],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[6,5],f32>) -> !torch.vtensor<[6,5],f32> 
     %12 = torch.operator "onnx.ReduceMean"(%11) {torch.onnx.axes = [1 : si64]} : (!torch.vtensor<[6,5],f32>) -> !torch.vtensor<[6,1],f32> 
@@ -29,8 +29,8 @@ module {
     %25 = torch.operator "onnx.Add"(%23, %24) : (!torch.vtensor<[6,5],f32>, !torch.vtensor<[1,5],f32>) -> !torch.vtensor<[6,5],f32> 
     %26 = torch.operator "onnx.Reshape"(%25, %2) : (!torch.vtensor<[6,5],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[2,3,5],f32> 
     %27 = torch.operator "onnx.Reciprocal"(%18) : (!torch.vtensor<[6,1],f32>) -> !torch.vtensor<[6,1],f32> 
-    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,3,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,3,1],f32> 
+    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[2,3,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[2,3,1],f32> 
     return %26, %28, %29 : !torch.vtensor<[2,3,5],f32>, !torch.vtensor<[2,3,1],f32>, !torch.vtensor<[2,3,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis2_epsilon_expanded_ver18/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis2_epsilon_expanded_ver18/model.mlir
index c17e7ea2d..fe5fdb7ed 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis2_epsilon_expanded_ver18/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_3d_axis2_epsilon_expanded_ver18/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<2> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[3],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[2],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[2],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[2],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[3],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[2,3,5],f32>) -> !torch.vtensor<[6,5],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[6,5],f32>) -> !torch.vtensor<[6,5],f32> 
     %12 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<1> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
@@ -30,8 +30,8 @@ module {
     %26 = torch.operator "onnx.Add"(%24, %25) : (!torch.vtensor<[6,5],f32>, !torch.vtensor<[1,5],f32>) -> !torch.vtensor<[6,5],f32> 
     %27 = torch.operator "onnx.Reshape"(%26, %2) : (!torch.vtensor<[6,5],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[2,3,5],f32> 
     %28 = torch.operator "onnx.Reciprocal"(%19) : (!torch.vtensor<[6,1],f32>) -> !torch.vtensor<[6,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,3,1],f32> 
-    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,3,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[2,3,1],f32> 
+    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[2,3,1],f32> 
     return %27, %29, %30 : !torch.vtensor<[2,3,5],f32>, !torch.vtensor<[2,3,1],f32>, !torch.vtensor<[2,3,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis0_expanded/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis0_expanded/model.mlir
index b0b775369..b826aa53f 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis0_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis0_expanded/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<0> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[4],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[0],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[0],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[4],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[0],si64>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[4],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[2,3,4,5],f32>) -> !torch.vtensor<[1,120],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[1,120],f32>) -> !torch.vtensor<[1,120],f32> 
     %12 = torch.operator "onnx.ReduceMean"(%11) {torch.onnx.axes = [1 : si64]} : (!torch.vtensor<[1,120],f32>) -> !torch.vtensor<[1,1],f32> 
@@ -29,8 +29,8 @@ module {
     %25 = torch.operator "onnx.Add"(%23, %24) : (!torch.vtensor<[1,120],f32>, !torch.vtensor<[1,120],f32>) -> !torch.vtensor<[1,120],f32> 
     %26 = torch.operator "onnx.Reshape"(%25, %2) : (!torch.vtensor<[1,120],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,4,5],f32> 
     %27 = torch.operator "onnx.Reciprocal"(%18) : (!torch.vtensor<[1,1],f32>) -> !torch.vtensor<[1,1],f32> 
-    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[1,1,1,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[1,1,1,1],f32> 
+    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[1,1,1,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[1,1,1,1],f32> 
     return %26, %28, %29 : !torch.vtensor<[2,3,4,5],f32>, !torch.vtensor<[1,1,1,1],f32>, !torch.vtensor<[1,1,1,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis0_expanded_ver18/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis0_expanded_ver18/model.mlir
index 77fca035f..e8384112d 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis0_expanded_ver18/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis0_expanded_ver18/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<0> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[4],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[0],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[0],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[4],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[0],si64>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[4],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[2,3,4,5],f32>) -> !torch.vtensor<[1,120],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[1,120],f32>) -> !torch.vtensor<[1,120],f32> 
     %12 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<1> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
@@ -30,8 +30,8 @@ module {
     %26 = torch.operator "onnx.Add"(%24, %25) : (!torch.vtensor<[1,120],f32>, !torch.vtensor<[1,120],f32>) -> !torch.vtensor<[1,120],f32> 
     %27 = torch.operator "onnx.Reshape"(%26, %2) : (!torch.vtensor<[1,120],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,4,5],f32> 
     %28 = torch.operator "onnx.Reciprocal"(%19) : (!torch.vtensor<[1,1],f32>) -> !torch.vtensor<[1,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[1,1,1,1],f32> 
-    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[1,1,1,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[1,1,1,1],f32> 
+    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[1,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[1,1,1,1],f32> 
     return %27, %29, %30 : !torch.vtensor<[2,3,4,5],f32>, !torch.vtensor<[1,1,1,1],f32>, !torch.vtensor<[1,1,1,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis1_expanded/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis1_expanded/model.mlir
index 45c5d829c..f83c9ecb3 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis1_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis1_expanded/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<1> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[4],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[1],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[3],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[1],si64>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[4],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 1 : si64} : (!torch.vtensor<[2,3,4,5],f32>) -> !torch.vtensor<[2,60],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[2,60],f32>) -> !torch.vtensor<[2,60],f32> 
     %12 = torch.operator "onnx.ReduceMean"(%11) {torch.onnx.axes = [1 : si64]} : (!torch.vtensor<[2,60],f32>) -> !torch.vtensor<[2,1],f32> 
@@ -29,8 +29,8 @@ module {
     %25 = torch.operator "onnx.Add"(%23, %24) : (!torch.vtensor<[2,60],f32>, !torch.vtensor<[1,60],f32>) -> !torch.vtensor<[2,60],f32> 
     %26 = torch.operator "onnx.Reshape"(%25, %2) : (!torch.vtensor<[2,60],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,4,5],f32> 
     %27 = torch.operator "onnx.Reciprocal"(%18) : (!torch.vtensor<[2,1],f32>) -> !torch.vtensor<[2,1],f32> 
-    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,1,1,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,1,1,1],f32> 
+    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,1,1,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,1,1,1],f32> 
     return %26, %28, %29 : !torch.vtensor<[2,3,4,5],f32>, !torch.vtensor<[2,1,1,1],f32>, !torch.vtensor<[2,1,1,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis1_expanded_ver18/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis1_expanded_ver18/model.mlir
index d459454b1..8853825be 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis1_expanded_ver18/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis1_expanded_ver18/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<1> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[4],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[1],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[3],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[1],si64>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[4],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 1 : si64} : (!torch.vtensor<[2,3,4,5],f32>) -> !torch.vtensor<[2,60],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[2,60],f32>) -> !torch.vtensor<[2,60],f32> 
     %12 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<1> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
@@ -30,8 +30,8 @@ module {
     %26 = torch.operator "onnx.Add"(%24, %25) : (!torch.vtensor<[2,60],f32>, !torch.vtensor<[1,60],f32>) -> !torch.vtensor<[2,60],f32> 
     %27 = torch.operator "onnx.Reshape"(%26, %2) : (!torch.vtensor<[2,60],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,4,5],f32> 
     %28 = torch.operator "onnx.Reciprocal"(%19) : (!torch.vtensor<[2,1],f32>) -> !torch.vtensor<[2,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,1,1,1],f32> 
-    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,1,1,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,1,1,1],f32> 
+    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[2,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,1,1,1],f32> 
     return %27, %29, %30 : !torch.vtensor<[2,3,4,5],f32>, !torch.vtensor<[2,1,1,1],f32>, !torch.vtensor<[2,1,1,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis2_expanded/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis2_expanded/model.mlir
index 87e1865fc..3994c6d7c 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis2_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis2_expanded/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<2> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[4],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[2],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[2],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[2],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[2],si64>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[4],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[2,3,4,5],f32>) -> !torch.vtensor<[6,20],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[6,20],f32>) -> !torch.vtensor<[6,20],f32> 
     %12 = torch.operator "onnx.ReduceMean"(%11) {torch.onnx.axes = [1 : si64]} : (!torch.vtensor<[6,20],f32>) -> !torch.vtensor<[6,1],f32> 
@@ -29,8 +29,8 @@ module {
     %25 = torch.operator "onnx.Add"(%23, %24) : (!torch.vtensor<[6,20],f32>, !torch.vtensor<[1,20],f32>) -> !torch.vtensor<[6,20],f32> 
     %26 = torch.operator "onnx.Reshape"(%25, %2) : (!torch.vtensor<[6,20],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,4,5],f32> 
     %27 = torch.operator "onnx.Reciprocal"(%18) : (!torch.vtensor<[6,1],f32>) -> !torch.vtensor<[6,1],f32> 
-    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,3,1,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,3,1,1],f32> 
+    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,1,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,1,1],f32> 
     return %26, %28, %29 : !torch.vtensor<[2,3,4,5],f32>, !torch.vtensor<[2,3,1,1],f32>, !torch.vtensor<[2,3,1,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis2_expanded_ver18/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis2_expanded_ver18/model.mlir
index 3bea71a68..9cf0ab084 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis2_expanded_ver18/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis2_expanded_ver18/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<2> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[4],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[2],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[2],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[2],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[2],si64>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[4],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[2,3,4,5],f32>) -> !torch.vtensor<[6,20],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[6,20],f32>) -> !torch.vtensor<[6,20],f32> 
     %12 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<1> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
@@ -30,8 +30,8 @@ module {
     %26 = torch.operator "onnx.Add"(%24, %25) : (!torch.vtensor<[6,20],f32>, !torch.vtensor<[1,20],f32>) -> !torch.vtensor<[6,20],f32> 
     %27 = torch.operator "onnx.Reshape"(%26, %2) : (!torch.vtensor<[6,20],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,4,5],f32> 
     %28 = torch.operator "onnx.Reciprocal"(%19) : (!torch.vtensor<[6,1],f32>) -> !torch.vtensor<[6,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,3,1,1],f32> 
-    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,3,1,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,1,1],f32> 
+    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[6,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,1,1],f32> 
     return %27, %29, %30 : !torch.vtensor<[2,3,4,5],f32>, !torch.vtensor<[2,3,1,1],f32>, !torch.vtensor<[2,3,1,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis3_expanded/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis3_expanded/model.mlir
index b7a1c04d6..b1319d340 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis3_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis3_expanded/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<3> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[4],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[3],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[3],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[3],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[4],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 3 : si64} : (!torch.vtensor<[2,3,4,5],f32>) -> !torch.vtensor<[24,5],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[24,5],f32>) -> !torch.vtensor<[24,5],f32> 
     %12 = torch.operator "onnx.ReduceMean"(%11) {torch.onnx.axes = [1 : si64]} : (!torch.vtensor<[24,5],f32>) -> !torch.vtensor<[24,1],f32> 
@@ -29,8 +29,8 @@ module {
     %25 = torch.operator "onnx.Add"(%23, %24) : (!torch.vtensor<[24,5],f32>, !torch.vtensor<[1,5],f32>) -> !torch.vtensor<[24,5],f32> 
     %26 = torch.operator "onnx.Reshape"(%25, %2) : (!torch.vtensor<[24,5],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,4,5],f32> 
     %27 = torch.operator "onnx.Reciprocal"(%18) : (!torch.vtensor<[24,1],f32>) -> !torch.vtensor<[24,1],f32> 
-    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[24,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,3,4,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[24,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,3,4,1],f32> 
+    %28 = torch.operator "onnx.Reshape"(%12, %9) : (!torch.vtensor<[24,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,4,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%27, %9) : (!torch.vtensor<[24,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,4,1],f32> 
     return %26, %28, %29 : !torch.vtensor<[2,3,4,5],f32>, !torch.vtensor<[2,3,4,1],f32>, !torch.vtensor<[2,3,4,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis3_expanded_ver18/model.mlir b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis3_expanded_ver18/model.mlir
index 00a481537..59365c826 100644
--- a/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis3_expanded_ver18/model.mlir
+++ b/iree_tests/onnx/node/generated/test_layer_normalization_4d_axis3_expanded_ver18/model.mlir
@@ -9,8 +9,8 @@ module {
     %5 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<3> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
     %6 = torch.operator "onnx.Slice"(%2, %4, %5) : (!torch.vtensor<[4],si64>, !torch.vtensor<[1],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[3],si64> 
     %7 = torch.operator "onnx.Sub"(%3, %5) : (!torch.vtensor<[],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
-    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[?],si64> 
-    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[3],si64>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[?],si64> 
+    %8 = torch.operator "onnx.ConstantOfShape"(%7) {torch.onnx.value = dense<1> : tensor<1xsi64>} : (!torch.vtensor<[1],si64>) -> !torch.vtensor<[1],si64> 
+    %9 = torch.operator "onnx.Concat"(%6, %8) {torch.onnx.axis = 0 : si64} : (!torch.vtensor<[3],si64>, !torch.vtensor<[1],si64>) -> !torch.vtensor<[4],si64> 
     %10 = torch.operator "onnx.Flatten"(%arg0) {torch.onnx.axis = 3 : si64} : (!torch.vtensor<[2,3,4,5],f32>) -> !torch.vtensor<[24,5],f32> 
     %11 = torch.operator "onnx.Cast"(%10) {torch.onnx.to = 1 : si64} : (!torch.vtensor<[24,5],f32>) -> !torch.vtensor<[24,5],f32> 
     %12 = torch.operator "onnx.Constant"() {torch.onnx.value = dense<1> : tensor<1xsi64>} : () -> !torch.vtensor<[1],si64> 
@@ -30,8 +30,8 @@ module {
     %26 = torch.operator "onnx.Add"(%24, %25) : (!torch.vtensor<[24,5],f32>, !torch.vtensor<[1,5],f32>) -> !torch.vtensor<[24,5],f32> 
     %27 = torch.operator "onnx.Reshape"(%26, %2) : (!torch.vtensor<[24,5],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,4,5],f32> 
     %28 = torch.operator "onnx.Reciprocal"(%19) : (!torch.vtensor<[24,1],f32>) -> !torch.vtensor<[24,1],f32> 
-    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[24,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,3,4,1],f32> 
-    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[24,1],f32>, !torch.vtensor<[?],si64>) -> !torch.vtensor<[2,3,4,1],f32> 
+    %29 = torch.operator "onnx.Reshape"(%13, %9) : (!torch.vtensor<[24,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,4,1],f32> 
+    %30 = torch.operator "onnx.Reshape"(%28, %9) : (!torch.vtensor<[24,1],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[2,3,4,1],f32> 
     return %27, %29, %30 : !torch.vtensor<[2,3,4,5],f32>, !torch.vtensor<[2,3,4,1],f32>, !torch.vtensor<[2,3,4,1],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_maxpool_2d_ceil_output_size_reduce_by_one/input_0.npy b/iree_tests/onnx/node/generated/test_maxpool_2d_ceil_output_size_reduce_by_one/input_0.npy
new file mode 100644
index 000000000..12db14385
Binary files /dev/null and b/iree_tests/onnx/node/generated/test_maxpool_2d_ceil_output_size_reduce_by_one/input_0.npy differ
diff --git a/iree_tests/onnx/node/generated/test_maxpool_2d_ceil_output_size_reduce_by_one/model.mlir b/iree_tests/onnx/node/generated/test_maxpool_2d_ceil_output_size_reduce_by_one/model.mlir
new file mode 100644
index 000000000..5409e9d45
--- /dev/null
+++ b/iree_tests/onnx/node/generated/test_maxpool_2d_ceil_output_size_reduce_by_one/model.mlir
@@ -0,0 +1,8 @@
+module {
+  func.func @test_maxpool_2d_ceil_output_size_reduce_by_one(%arg0: !torch.vtensor<[1,1,2,2],f32>) -> !torch.vtensor<[1,1,1,1],f32> attributes {torch.onnx_meta.ir_version = 7 : si64, torch.onnx_meta.opset_version = 17 : si64, torch.onnx_meta.producer_name = "backend-test", torch.onnx_meta.producer_version = ""} {
+    %none = torch.constant.none
+    %0 = torch.operator "onnx.MaxPool"(%arg0) {torch.onnx.ceil_mode = 1 : si64, torch.onnx.kernel_shape = [1 : si64, 1 : si64], torch.onnx.strides = [2 : si64, 2 : si64]} : (!torch.vtensor<[1,1,2,2],f32>) -> !torch.vtensor<[1,1,1,1],f32> 
+    return %0 : !torch.vtensor<[1,1,1,1],f32>
+  }
+}
+
diff --git a/iree_tests/onnx/node/generated/test_maxpool_2d_ceil_output_size_reduce_by_one/output_0.npy b/iree_tests/onnx/node/generated/test_maxpool_2d_ceil_output_size_reduce_by_one/output_0.npy
new file mode 100644
index 000000000..f57858a92
Binary files /dev/null and b/iree_tests/onnx/node/generated/test_maxpool_2d_ceil_output_size_reduce_by_one/output_0.npy differ
diff --git a/iree_tests/onnx/node/generated/test_maxpool_2d_ceil_output_size_reduce_by_one/test_data_flags.txt b/iree_tests/onnx/node/generated/test_maxpool_2d_ceil_output_size_reduce_by_one/test_data_flags.txt
new file mode 100644
index 000000000..1eca09b71
--- /dev/null
+++ b/iree_tests/onnx/node/generated/test_maxpool_2d_ceil_output_size_reduce_by_one/test_data_flags.txt
@@ -0,0 +1,2 @@
+--input=1x1x2x2xf32=@input_0.bin
+--expected_output=1x1x1x1xf32=@output_0.bin
diff --git a/iree_tests/onnx/node/generated/test_sce_NCd1_mean_weight_negative_ii_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_NCd1_mean_weight_negative_ii_expanded/model.mlir
index 4b36e75ab..7de38a6f2 100644
--- a/iree_tests/onnx/node/generated/test_sce_NCd1_mean_weight_negative_ii_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_NCd1_mean_weight_negative_ii_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,6,5],f32>) -> !torch.vtensor<[3,6,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,6,5],f32>) -> !torch.vtensor<[3,5,6],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,6],f32>) -> !torch.vtensor<[3],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,6],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[?,?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = -1 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[3,6],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,6],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3,5,6],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = -1 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5,6],f32>, !torch.vtensor<[3,6],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
     return %7 : !torch.vtensor<[],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_NCd1_mean_weight_negative_ii_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_NCd1_mean_weight_negative_ii_log_prob_expanded/model.mlir
index ee90b011a..705ab4393 100644
--- a/iree_tests/onnx/node/generated/test_sce_NCd1_mean_weight_negative_ii_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_NCd1_mean_weight_negative_ii_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,6,5],f32>) -> !torch.vtensor<[3,6,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,6,5],f32>) -> !torch.vtensor<[3,5,6],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,6],f32>) -> !torch.vtensor<[3],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,6],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[?,?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?,?],f32>) -> !torch.vtensor<[3,5,6],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = -1 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[3,6],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,6],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3,5,6],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5,6],f32>) -> !torch.vtensor<[3,5,6],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = -1 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5,6],f32>, !torch.vtensor<[3,6],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
     return %8, %7 : !torch.vtensor<[],f32>, !torch.vtensor<[3,5,6],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_none_no_weight_negative_ii_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_none_no_weight_negative_ii_expanded/model.mlir
index c823f6073..bc31e5154 100644
--- a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_none_no_weight_negative_ii_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_none_no_weight_negative_ii_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,180,5],f32>) -> !torch.vtensor<[3,180,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,180,5],f32>) -> !torch.vtensor<[3,5,180],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,6,6,5],f32>) -> !torch.vtensor<[5],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,180],f32>, !torch.vtensor<[5],si64>) -> !torch.vtensor<[?,?,?,?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = -5 : si64, torch.onnx.reduction = "none"} : (!torch.vtensor<[?,?,?,?,?],f32>, !torch.vtensor<[3,6,6,5],si64>) -> !torch.vtensor<[3,6,6,5],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,180],f32>, !torch.vtensor<[5],si64>) -> !torch.vtensor<[3,5,6,6,5],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = -5 : si64, torch.onnx.reduction = "none"} : (!torch.vtensor<[3,5,6,6,5],f32>, !torch.vtensor<[3,6,6,5],si64>) -> !torch.vtensor<[3,6,6,5],f32> 
     return %7 : !torch.vtensor<[3,6,6,5],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_none_no_weight_negative_ii_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_none_no_weight_negative_ii_log_prob_expanded/model.mlir
index d1ee8d7a1..a3050746b 100644
--- a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_none_no_weight_negative_ii_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_none_no_weight_negative_ii_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,180,5],f32>) -> !torch.vtensor<[3,180,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,180,5],f32>) -> !torch.vtensor<[3,5,180],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,6,6,5],f32>) -> !torch.vtensor<[5],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,180],f32>, !torch.vtensor<[5],si64>) -> !torch.vtensor<[?,?,?,?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?,?,?,?],f32>) -> !torch.vtensor<[3,5,6,6,5],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = -5 : si64, torch.onnx.reduction = "none"} : (!torch.vtensor<[?,?,?,?,?],f32>, !torch.vtensor<[3,6,6,5],si64>) -> !torch.vtensor<[3,6,6,5],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,180],f32>, !torch.vtensor<[5],si64>) -> !torch.vtensor<[3,5,6,6,5],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5,6,6,5],f32>) -> !torch.vtensor<[3,5,6,6,5],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = -5 : si64, torch.onnx.reduction = "none"} : (!torch.vtensor<[3,5,6,6,5],f32>, !torch.vtensor<[3,6,6,5],si64>) -> !torch.vtensor<[3,6,6,5],f32> 
     return %8, %7 : !torch.vtensor<[3,6,6,5],f32>, !torch.vtensor<[3,5,6,6,5],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_sum_weight_high_ii_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_sum_weight_high_ii_expanded/model.mlir
index 49e3e596e..23202545f 100644
--- a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_sum_weight_high_ii_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_sum_weight_high_ii_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 10 : si64, torch.onnx.reduction = "sum"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 10 : si64, torch.onnx.reduction = "sum"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
     return %7 : !torch.vtensor<[],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_sum_weight_high_ii_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_sum_weight_high_ii_log_prob_expanded/model.mlir
index 3f4051e5c..11e7d39ef 100644
--- a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_sum_weight_high_ii_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3_sum_weight_high_ii_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?],f32>) -> !torch.vtensor<[3,5],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 10 : si64, torch.onnx.reduction = "sum"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[3,5],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 10 : si64, torch.onnx.reduction = "sum"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
     return %8, %7 : !torch.vtensor<[],f32>, !torch.vtensor<[3,5],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_mean_weight_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_mean_weight_expanded/model.mlir
index 2fbe9d1b9..39f49710b 100644
--- a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_mean_weight_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_mean_weight_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,2160,5],f32>) -> !torch.vtensor<[3,2160,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,2160,5],f32>) -> !torch.vtensor<[3,5,2160],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,6,6,5,3,4],f32>) -> !torch.vtensor<[7],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2160],f32>, !torch.vtensor<[7],si64>) -> !torch.vtensor<[?,?,?,?,?,?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?,?,?,?,?,?],f32>, !torch.vtensor<[3,6,6,5,3,4],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2160],f32>, !torch.vtensor<[7],si64>) -> !torch.vtensor<[3,5,6,6,5,3,4],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5,6,6,5,3,4],f32>, !torch.vtensor<[3,6,6,5,3,4],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
     return %7 : !torch.vtensor<[],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_mean_weight_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_mean_weight_log_prob_expanded/model.mlir
index 416d8a00a..bdaa3abc3 100644
--- a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_mean_weight_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_mean_weight_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,2160,5],f32>) -> !torch.vtensor<[3,2160,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,2160,5],f32>) -> !torch.vtensor<[3,5,2160],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,6,6,5,3,4],f32>) -> !torch.vtensor<[7],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2160],f32>, !torch.vtensor<[7],si64>) -> !torch.vtensor<[?,?,?,?,?,?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?,?,?,?,?,?],f32>) -> !torch.vtensor<[3,5,6,6,5,3,4],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?,?,?,?,?,?],f32>, !torch.vtensor<[3,6,6,5,3,4],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2160],f32>, !torch.vtensor<[7],si64>) -> !torch.vtensor<[3,5,6,6,5,3,4],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5,6,6,5,3,4],f32>) -> !torch.vtensor<[3,5,6,6,5,3,4],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5,6,6,5,3,4],f32>, !torch.vtensor<[3,6,6,5,3,4],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
     return %8, %7 : !torch.vtensor<[],f32>, !torch.vtensor<[3,5,6,6,5,3,4],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_none_no_weight_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_none_no_weight_expanded/model.mlir
index b0901eb9a..749e6347d 100644
--- a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_none_no_weight_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_none_no_weight_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,2160,5],f32>) -> !torch.vtensor<[3,2160,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,2160,5],f32>) -> !torch.vtensor<[3,5,2160],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,6,6,5,3,4],f32>) -> !torch.vtensor<[7],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2160],f32>, !torch.vtensor<[7],si64>) -> !torch.vtensor<[?,?,?,?,?,?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "none"} : (!torch.vtensor<[?,?,?,?,?,?,?],f32>, !torch.vtensor<[3,6,6,5,3,4],si64>) -> !torch.vtensor<[3,6,6,5,3,4],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2160],f32>, !torch.vtensor<[7],si64>) -> !torch.vtensor<[3,5,6,6,5,3,4],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "none"} : (!torch.vtensor<[3,5,6,6,5,3,4],f32>, !torch.vtensor<[3,6,6,5,3,4],si64>) -> !torch.vtensor<[3,6,6,5,3,4],f32> 
     return %7 : !torch.vtensor<[3,6,6,5,3,4],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_none_no_weight_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_none_no_weight_log_prob_expanded/model.mlir
index fd899d1d2..e87e26b32 100644
--- a/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_none_no_weight_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_NCd1d2d3d4d5_none_no_weight_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,2160,5],f32>) -> !torch.vtensor<[3,2160,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,2160,5],f32>) -> !torch.vtensor<[3,5,2160],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,6,6,5,3,4],f32>) -> !torch.vtensor<[7],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2160],f32>, !torch.vtensor<[7],si64>) -> !torch.vtensor<[?,?,?,?,?,?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?,?,?,?,?,?],f32>) -> !torch.vtensor<[3,5,6,6,5,3,4],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "none"} : (!torch.vtensor<[?,?,?,?,?,?,?],f32>, !torch.vtensor<[3,6,6,5,3,4],si64>) -> !torch.vtensor<[3,6,6,5,3,4],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2160],f32>, !torch.vtensor<[7],si64>) -> !torch.vtensor<[3,5,6,6,5,3,4],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5,6,6,5,3,4],f32>) -> !torch.vtensor<[3,5,6,6,5,3,4],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "none"} : (!torch.vtensor<[3,5,6,6,5,3,4],f32>, !torch.vtensor<[3,6,6,5,3,4],si64>) -> !torch.vtensor<[3,6,6,5,3,4],f32> 
     return %8, %7 : !torch.vtensor<[3,6,6,5,3,4],f32>, !torch.vtensor<[3,5,6,6,5,3,4],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_3d_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_3d_expanded/model.mlir
index 628ba1c0a..a05668fee 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_3d_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_3d_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,2,5],f32>) -> !torch.vtensor<[3,2,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,2,5],f32>) -> !torch.vtensor<[3,5,2],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,2],f32>) -> !torch.vtensor<[3],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[?,?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[3,2],si64>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3,5,2],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3,2],si64>) -> !torch.vtensor<[],f32> 
     return %7 : !torch.vtensor<[],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_3d_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_3d_log_prob_expanded/model.mlir
index af4251bf8..ae68b7f9a 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_3d_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_3d_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,2,5],f32>) -> !torch.vtensor<[3,2,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,2,5],f32>) -> !torch.vtensor<[3,5,2],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,2],f32>) -> !torch.vtensor<[3],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[?,?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?,?],f32>) -> !torch.vtensor<[3,5,2],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[3,2],si64>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3,5,2],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5,2],f32>) -> !torch.vtensor<[3,5,2],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3,2],si64>) -> !torch.vtensor<[],f32> 
     return %8, %7 : !torch.vtensor<[],f32>, !torch.vtensor<[3,5,2],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_expanded/model.mlir
index a83cc8f85..93567a507 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[],f32> 
     return %7 : !torch.vtensor<[],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_log_prob_expanded/model.mlir
index a6c244317..f0ffcdc75 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?],f32>) -> !torch.vtensor<[3,5],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[3,5],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[],f32> 
     return %8, %7 : !torch.vtensor<[],f32>, !torch.vtensor<[3,5],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_3d_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_3d_expanded/model.mlir
index c386761f1..406a38976 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_3d_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_3d_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,2,5],f32>) -> !torch.vtensor<[3,2,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,2,5],f32>) -> !torch.vtensor<[3,5,2],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,2],f32>) -> !torch.vtensor<[3],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[?,?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[3,2],si64>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3,5,2],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3,2],si64>) -> !torch.vtensor<[],f32> 
     return %7 : !torch.vtensor<[],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_3d_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_3d_log_prob_expanded/model.mlir
index 389d7ade9..29d90b6a1 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_3d_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_3d_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,2,5],f32>) -> !torch.vtensor<[3,2,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,2,5],f32>) -> !torch.vtensor<[3,5,2],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,2],f32>) -> !torch.vtensor<[3],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[?,?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?,?],f32>) -> !torch.vtensor<[3,5,2],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[3,2],si64>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3,5,2],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5,2],f32>) -> !torch.vtensor<[3,5,2],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3,2],si64>) -> !torch.vtensor<[],f32> 
     return %8, %7 : !torch.vtensor<[],f32>, !torch.vtensor<[3,5,2],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_4d_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_4d_expanded/model.mlir
index 41b3609aa..401c7849b 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_4d_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_4d_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,14,5],f32>) -> !torch.vtensor<[3,14,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,14,5],f32>) -> !torch.vtensor<[3,5,14],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,2,7],f32>) -> !torch.vtensor<[4],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,14],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[?,?,?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?,?,?],f32>, !torch.vtensor<[3,2,7],si64>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,14],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[3,5,2,7],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5,2,7],f32>, !torch.vtensor<[3,2,7],si64>) -> !torch.vtensor<[],f32> 
     return %7 : !torch.vtensor<[],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_4d_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_4d_log_prob_expanded/model.mlir
index 36777c352..64540aaf8 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_4d_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_4d_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,14,5],f32>) -> !torch.vtensor<[3,14,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,14,5],f32>) -> !torch.vtensor<[3,5,14],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,2,7],f32>) -> !torch.vtensor<[4],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,14],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[?,?,?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?,?,?],f32>) -> !torch.vtensor<[3,5,2,7],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?,?,?],f32>, !torch.vtensor<[3,2,7],si64>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,14],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[3,5,2,7],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5,2,7],f32>) -> !torch.vtensor<[3,5,2,7],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5,2,7],f32>, !torch.vtensor<[3,2,7],si64>) -> !torch.vtensor<[],f32> 
     return %8, %7 : !torch.vtensor<[],f32>, !torch.vtensor<[3,5,2,7],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_expanded/model.mlir
index 3941b9348..f85040dd9 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[],f32> 
     return %7 : !torch.vtensor<[],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_log_prob_expanded/model.mlir
index 92deb6ae7..5964d3c35 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_no_weight_ii_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?],f32>) -> !torch.vtensor<[3,5],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[3,5],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[],f32> 
     return %8, %7 : !torch.vtensor<[],f32>, !torch.vtensor<[3,5],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_weight_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_weight_expanded/model.mlir
index a6a170e94..811c29cf6 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_weight_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_weight_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
     return %7 : !torch.vtensor<[],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_3d_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_3d_expanded/model.mlir
index 60d507596..3245b8d47 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_3d_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_3d_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,2,5],f32>) -> !torch.vtensor<[3,2,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,2,5],f32>) -> !torch.vtensor<[3,5,2],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,2],f32>) -> !torch.vtensor<[3],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[?,?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 1 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[3,2],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3,5,2],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 1 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3,2],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
     return %7 : !torch.vtensor<[],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_3d_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_3d_log_prob_expanded/model.mlir
index ad3ca3dab..6c6a3dfbe 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_3d_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_3d_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,2,5],f32>) -> !torch.vtensor<[3,2,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,2,5],f32>) -> !torch.vtensor<[3,5,2],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,2],f32>) -> !torch.vtensor<[3],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[?,?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?,?],f32>) -> !torch.vtensor<[3,5,2],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 1 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?,?],f32>, !torch.vtensor<[3,2],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3,5,2],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5,2],f32>) -> !torch.vtensor<[3,5,2],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 1 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5,2],f32>, !torch.vtensor<[3,2],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
     return %8, %7 : !torch.vtensor<[],f32>, !torch.vtensor<[3,5,2],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_4d_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_4d_expanded/model.mlir
index a90d45982..277fe9a35 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_4d_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_4d_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,14,5],f32>) -> !torch.vtensor<[3,14,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,14,5],f32>) -> !torch.vtensor<[3,5,14],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,2,7],f32>) -> !torch.vtensor<[4],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,14],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[?,?,?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?,?,?],f32>, !torch.vtensor<[3,2,7],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,14],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[3,5,2,7],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5,2,7],f32>, !torch.vtensor<[3,2,7],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
     return %7 : !torch.vtensor<[],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_4d_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_4d_log_prob_expanded/model.mlir
index 88fb63cc3..2943fd2f1 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_4d_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_4d_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,14,5],f32>) -> !torch.vtensor<[3,14,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,14,5],f32>) -> !torch.vtensor<[3,5,14],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5,2,7],f32>) -> !torch.vtensor<[4],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,14],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[?,?,?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?,?,?],f32>) -> !torch.vtensor<[3,5,2,7],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?,?,?],f32>, !torch.vtensor<[3,2,7],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,14],f32>, !torch.vtensor<[4],si64>) -> !torch.vtensor<[3,5,2,7],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5,2,7],f32>) -> !torch.vtensor<[3,5,2,7],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 2 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5,2,7],f32>, !torch.vtensor<[3,2,7],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
     return %8, %7 : !torch.vtensor<[],f32>, !torch.vtensor<[3,5,2,7],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_expanded/model.mlir
index 5730c72d9..e04187d92 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 0 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 0 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
     return %7 : !torch.vtensor<[],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_log_prob_expanded/model.mlir
index e5fd9b2e1..f7e9a2196 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_weight_ii_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?],f32>) -> !torch.vtensor<[3,5],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 0 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[3,5],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.ignore_index = 0 : si64, torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
     return %8, %7 : !torch.vtensor<[],f32>, !torch.vtensor<[3,5],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_mean_weight_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_mean_weight_log_prob_expanded/model.mlir
index 08d6b4638..8aab8b1ac 100644
--- a/iree_tests/onnx/node/generated/test_sce_mean_weight_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_mean_weight_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?],f32>) -> !torch.vtensor<[3,5],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[3,5],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.reduction = "mean"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[],f32> 
     return %8, %7 : !torch.vtensor<[],f32>, !torch.vtensor<[3,5],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_none_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_none_expanded/model.mlir
index 238f367ad..cc866c2ed 100644
--- a/iree_tests/onnx/node/generated/test_sce_none_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_none_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "none"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "none"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3],f32> 
     return %7 : !torch.vtensor<[3],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_none_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_none_log_prob_expanded/model.mlir
index 8ea5da4ed..570a7d08f 100644
--- a/iree_tests/onnx/node/generated/test_sce_none_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_none_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?],f32>) -> !torch.vtensor<[3,5],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "none"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[3,5],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "none"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[3],f32> 
     return %8, %7 : !torch.vtensor<[3],f32>, !torch.vtensor<[3,5],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_none_weights_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_none_weights_expanded/model.mlir
index ce900ac8c..7692278f7 100644
--- a/iree_tests/onnx/node/generated/test_sce_none_weights_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_none_weights_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.reduction = "none"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[3],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.reduction = "none"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[3],f32> 
     return %7 : !torch.vtensor<[3],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_none_weights_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_none_weights_log_prob_expanded/model.mlir
index 8e726d747..3ebb3d0ea 100644
--- a/iree_tests/onnx/node/generated/test_sce_none_weights_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_none_weights_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?],f32>) -> !torch.vtensor<[3,5],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.reduction = "none"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[3],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[3,5],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1, %arg2) {torch.onnx.reduction = "none"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>, !torch.vtensor<[5],f32>) -> !torch.vtensor<[3],f32> 
     return %8, %7 : !torch.vtensor<[3],f32>, !torch.vtensor<[3,5],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_sum_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_sum_expanded/model.mlir
index 7604ba4e0..b9a1cacda 100644
--- a/iree_tests/onnx/node/generated/test_sce_sum_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_sum_expanded/model.mlir
@@ -7,8 +7,8 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "sum"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "sum"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[],f32> 
     return %7 : !torch.vtensor<[],f32>
   }
 }
diff --git a/iree_tests/onnx/node/generated/test_sce_sum_log_prob_expanded/model.mlir b/iree_tests/onnx/node/generated/test_sce_sum_log_prob_expanded/model.mlir
index 23c547841..586cfacde 100644
--- a/iree_tests/onnx/node/generated/test_sce_sum_log_prob_expanded/model.mlir
+++ b/iree_tests/onnx/node/generated/test_sce_sum_log_prob_expanded/model.mlir
@@ -7,9 +7,9 @@ module {
     %3 = torch.operator "onnx.LogSoftmax"(%2) {torch.onnx.axis = 2 : si64} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,1,5],f32> 
     %4 = torch.operator "onnx.Transpose"(%3) {torch.onnx.perm = [0 : si64, 2 : si64, 1 : si64]} : (!torch.vtensor<[3,1,5],f32>) -> !torch.vtensor<[3,5,1],f32> 
     %5 = torch.operator "onnx.Shape"(%arg0) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[2],si64> 
-    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[?,?],f32> 
-    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[?,?],f32>) -> !torch.vtensor<[3,5],f32> 
-    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "sum"} : (!torch.vtensor<[?,?],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[],f32> 
+    %6 = torch.operator "onnx.Reshape"(%4, %5) : (!torch.vtensor<[3,5,1],f32>, !torch.vtensor<[2],si64>) -> !torch.vtensor<[3,5],f32> 
+    %7 = torch.operator "onnx.Identity"(%6) : (!torch.vtensor<[3,5],f32>) -> !torch.vtensor<[3,5],f32> 
+    %8 = torch.operator "onnx.NegativeLogLikelihoodLoss"(%6, %arg1) {torch.onnx.reduction = "sum"} : (!torch.vtensor<[3,5],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[],f32> 
     return %8, %7 : !torch.vtensor<[],f32>, !torch.vtensor<[3,5],f32>
   }
 }
diff --git a/iree_tests/onnx/node/import_failures.txt b/iree_tests/onnx/node/import_failures.txt
index ddd750b5d..c1599c770 100644
--- a/iree_tests/onnx/node/import_failures.txt
+++ b/iree_tests/onnx/node/import_failures.txt
@@ -11,7 +11,6 @@ test_cast_INT4_to_INT8
 test_cast_UINT4_to_FLOAT
 test_cast_UINT4_to_FLOAT16
 test_cast_UINT4_to_UINT8
-test_dequantizelinear_e4m3fn_float16
 test_dequantizelinear_int4
 test_dequantizelinear_uint4
 test_identity_opt
@@ -20,7 +19,6 @@ test_if_opt
 test_if_seq
 test_loop13_seq
 test_loop16_seq_none
-test_maxpool_2d_ceil_output_size_reduce_by_one
 test_optional_get_element_optional_sequence
 test_optional_get_element_optional_tensor
 test_optional_get_element_sequence
diff --git a/iree_tests/onnx/node/import_successes.txt b/iree_tests/onnx/node/import_successes.txt
index 4b6a8ab5f..d1bee9e2c 100644
--- a/iree_tests/onnx/node/import_successes.txt
+++ b/iree_tests/onnx/node/import_successes.txt
@@ -301,6 +301,7 @@ test_dequantizelinear
 test_dequantizelinear_axis
 test_dequantizelinear_blocked
 test_dequantizelinear_e4m3fn
+test_dequantizelinear_e4m3fn_float16
 test_dequantizelinear_e4m3fn_zero_point
 test_dequantizelinear_e5m2
 test_dequantizelinear_int16
@@ -598,6 +599,7 @@ test_max_uint64
 test_max_uint8
 test_maxpool_1d_default
 test_maxpool_2d_ceil
+test_maxpool_2d_ceil_output_size_reduce_by_one
 test_maxpool_2d_default
 test_maxpool_2d_dilations
 test_maxpool_2d_pads