AmosLewis

import torch


def nonzero(t):
    print("t: ", t)  # tensor([0, 0, 1, 1, 0, 0])
    # Flatten the input tensor
    t_flat = t.flatten()  #  torch.flatten(t, 0, 0)
    print(
        "t_flat: ", t_flat
    )  # tensortensor([0, 0, 1, 1, 0, 0]), torch.Size([6]), #!torch.vtensor<[?],si64>

AmosLewis

iree-compile --iree-hal-target-backends=llvm-cpu model.linalg.mlir -o model.vmfb --dump-compilation-phases-to=./tmp/
failed to translate executables
failed to translate executables
model.linalg.mlir:21:10: error: 'memref.alloca' op expected no unbounded stack allocations
    %1 = tensor.empty(%dim) : tensor<?xi64>
         ^
model.linalg.mlir:10:3: note: called from
  func.func @main_graph(%arg0: tensor<?xi1>) -> tensor<1x1xi64> {
  ^
model.linalg.mlir:21:10: note: see current operation: %14 = "memref.alloca"(%11) <{alignment = 64 : i64, operandSegmentSizes = array<i32: 1, 0>}> : (index) -> memref<?xi64>

AmosLewis

Running MaskedFillTensorIntValueStaticModule_basic...
*** RUNNING TEST: MaskedScatterStaticBasic_basic ***
Compiling MaskedScatterStaticBasic_basic...
/proj/gdba/shark/chi/src/torch-mlir/mlir_venv/lib/python3.10/site-packages/torch/onnx/symbolic_opset10.py:513: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
  return g.op("Constant", value_t=torch.tensor(list_or_value))

====================
ONNX RAW IR
module {
  func.func @main_graph(%arg0: !torch.vtensor<[4,4],f32>, %arg1: !torch.vtensor<[4,4],i1>, %arg2: !torch.vtensor<[8,8],f32>) -> !torch.vtensor<[4,4],f32> attributes {torch.onnx_meta.ir_version = 9 : si64, torch.onnx_meta.opset_version = 20 : si64, torch.onnx_meta.producer_name = "pytorch", torch.onnx_meta.producer_version = "2.6.0"} {

AmosLewis

./build_tools/ci/test_posix.sh
::group::Run ONNX e2e integration tests
TORCH_VERSION_FOR_COMPARISON = 2.6.0.dev20241107
Running tests sequentially with progress status
*** RUNNING TEST: AtenNonzero1DModule_one_nonzero ***
Compiling AtenNonzero1DModule_one_nonzero...

====================
ONNX RAW IR
module {

AmosLewis

module {
  func.func @tf2onnx(%arg0: !torch.vtensor<[?,768],f32>, %arg1: !torch.vtensor<[3],si64>, %arg2: !torch.vtensor<[?,256,768],f32>) -> ( !torch.vtensor<[?,256,768],f32>) attributes {torch.onnx_meta.ir_version = 7 : si64, torch.onnx_meta.opset_version = 21 : si64, torch.onnx_meta.producer_name = "tf2onnx", torch.onnx_meta.producer_version = "1.5.2"} {
    %reshape = torch.operator "onnx.Reshape"(%arg0, %arg1) : (!torch.vtensor<[?,768],f32>, !torch.vtensor<[3],si64>) -> !torch.vtensor<[?,256,768],f32> 
    %866 = torch.operator "onnx.Add"(%reshape, %arg2) : (!torch.vtensor<[?,256,768],f32>, !torch.vtensor<[?,256,768],f32>) -> !torch.vtensor<[?,256,768],f32> 
    return %866 :  !torch.vtensor<[?,256,768],f32>
  }
}

AmosLewis

<unknown>:0: error: failed to legalize unresolved materialization from ('i64') to 'index' that remained live after conversion
<unknown>:0: note: see current operation: %452 = "builtin.unrealized_conversion_cast"(%451) : (i64) -> index
/proj/gdba/shark/chi/src/SHARK-TestSuite/alt_e2eshark/test-run/model--long-t5-tglobal-base-16384-book-summary--pszemraj/model.modified.mlir:1299:12: note: see existing live user here: %2529 = hal.interface.binding.subspan layout(<constants = 20, bindings = [#hal.pipeline.binding<storage_buffer, "ReadOnly|Indirect">, #hal.pipeline.binding<storage_buffer, "ReadOnly|Indirect">, #hal.pipeline.binding<storage_buffer, "ReadOnly|Indirect">, #hal.pipeline.binding<storage_buffer, Indirect>], flags = Indirect>) binding(0) alignment(64) offset(%3) flags("ReadOnly|Indirect") : memref<i64, strided<[], offset: 24>>
    %360 = linalg.generic {indexing_maps = [#map12, #map12, #map12, #map12], iterator_types = []} ins(%357, %358, %359 : tensor<i1>, tensor<i64>, tensor<i64>) outs(%94 : tensor<i64

AmosLewis

module {
  func.func @main_graph(%arg5:!torch.vtensor<[2708],f32>, %arg1: !torch.vtensor<[?],si64>, %arg2: !torch.vtensor<[?],f32>) -> !torch.vtensor<[2708],f32>    attributes {torch.onnx_meta.ir_version = 8 : si64, torch.onnx_meta.opset_version = 17 : si64, torch.onnx_meta.producer_name = "pytorch", torch.onnx_meta.producer_version = "2.1.0"} {
    %59 = torch.operator "onnx.ScatterElements"(%arg5, %arg1, %arg2) {torch.onnx.axis = 0 : si64, torch.onnx.reduction = "add"} : (!torch.vtensor<[2708],f32>, !torch.vtensor<[?],si64>, !torch.vtensor<[?],f32>) -> !torch.vtensor<[2708],f32>
    return %59 : !torch.vtensor<[2708],f32>
  }
}

AmosLewis

python ./run.py  --tolerance 0.001 0.001 --cachedir /proj/gdba/shark/cache --ireebuild ../../iree-build -f onnx -g models --mode onnx --report -j 12  -r test-ru
n-vision_int8 --testsfile list1_vision_int8_run
/proj/gdba/shark/chi/src/SHARK-TestSuite/e2eshark/e2e_venv/lib/python3.10/site-packages/torchvision/io/image.py:14: UserWarning: Failed to load image Python extension: '/proj/gdba/shark/chi/src/SHARK-TestSuite/e2eshark/e2e_venv/lib/python3.10/site-packages/torchvision/image.so: undefined symbol: _ZNK3c1011StorageImpl27throw_data_ptr_access_errorEv'If you don't plan on using image functionality from `torchvision.io`, you can ignore this warning. Otherwise, there might be something wrong with your environment. Did you have `libjpeg` or `libpng` installed before building `torchvision` from source?
  warn(
Starting e2eshark tests. Using 12 processes
Cache Directory: /proj/gdba/shark/cache
Tolerance for comparing floating point (atol, rtol) = (0.001, 0.001)
Note: No Torch MLIR build provided using --torchmlir

AmosLewis

hal.executable public @torch_jit_dispatch_33 {
  hal.executable.variant public @embedded_elf_x86_64 target(<"llvm-cpu", "embedded-elf-x86_64", {cpu = "generic", cpu_features = "", data_layout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128", native_vector_size = 16 : i64, target_triple = "x86_64-unknown-unknown-eabi-elf"}>) {
    hal.executable.export public @torch_jit_dispatch_33_quantized_batch_matmul_56x56x512x128_i8xi8xi32xi32xi32 ordinal(0) layout(#hal.pipeline.layout<push_constants = 0, sets = [<0, bindings = [<0, storage_buffer, ReadOnly>, <1, storage_buffer, ReadOnly>, <2, storage_buffer>], flags = Indirect>]>) attributes {hal.interface.bindings = [#hal.interface.binding<0, 0>, #hal.interface.binding<0, 1>, #hal.interface.binding<0, 2>]} {
    ^bb0(%arg0: !hal.device):
      %x, %y, %z = flow.dispatch.workgroup_count_from_slice 
      hal.return %x, %y, %z : index, index, index
    }
    builtin.module {
      func.func @torch_jit_dispatch_33_quantized_batch_matmul_5

AmosLewis

hal.executable public @torch_jit_dispatch_25 {
  hal.executable.variant public @embedded_elf_x86_64 target(<"llvm-cpu", "embedded-elf-x86_64", {cpu = "generic", cpu_features = "", data_layout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128", native_vector_size = 16 : i64, target_triple = "x86_64-unknown-unknown-eabi-elf"}>) {
    hal.executable.export public @torch_jit_dispatch_25_quantized_batch_matmul_56x56x128x512_i8xi8xi32xi32xi32 ordinal(0) layout(#hal.pipeline.layout<push_constants = 0, sets = [<0, bindings = [<0, storage_buffer, ReadOnly>, <1, storage_buffer, ReadOnly>, <2, storage_buffer>], flags = Indirect>]>) attributes {hal.interface.bindings = [#hal.interface.binding<0, 0>, #hal.interface.binding<0, 1>, #hal.interface.binding<0, 2>]} {
    ^bb0(%arg0: !hal.device):
      %x, %y, %z = flow.dispatch.workgroup_count_from_slice 
      hal.return %x, %y, %z : index, index, index
    }
    builtin.module {
      func.func @torch_jit_dispatch_25_quantized_batch_matmul_5