renxida

OVERVIEW: IREE compilation driver

USAGE: iree-compile [options] <input file or '-' for stdin>

OPTIONS:

CUDA HAL Target:

  --iree-hal-cuda-dump-ptx                                                   - Dump ptx to the debug stream.
  --iree-hal-cuda-llvm-target-arch=<string>                                  - LLVM target chip.

renxida

We suffer more in imagination than in reality - - Seneca

Situation / Action

Right now

Define fears

• fear:
  • prevent by:
  • repair by:
• fear:

renxida

 ~/torch-mlir/build/bin/torch-mlir-opt --convert-torch-to-linalg --convert-torch-to-tmtensor --debug -mlir-disable-threading -mlir-print-ir-after-all ./stripped-opt-125M.fp32.onnx.torch.mlir &> /tmp/torchopt.out

/home/azureuser/torch-mlir/build/bin/torch-mlir-opt: /home/azureuser/miniconda/lib/libtinfo.so.6: no version information available (required by /home/azureuser/torch-mlir/build/bin/torch-mlir-opt)
Args: /home/azureuser/torch-mlir/build/bin/torch-mlir-opt --convert-torch-to-linalg --convert-torch-to-tmtensor --debug -mlir-disable-threading -mlir-print-ir-after-all ./stripped-opt-125M.fp32.onnx.torch.mlir 
Load new dialect in Context builtin
ImplicitTypeIDRegistry::lookupOrInsert(mlir::ShapedType)
ImplicitTypeIDRegistry::lookupOrInsert(mlir::MemRefLayoutAttrInterface)
ImplicitTypeIDRegistry::lookupOrInsert(mlir::TypedAttr)
ImplicitTypeIDRegistry::lookupOrInsert(mlir::ElementsAttr)
ImplicitTypeIDRegistry::lookupOrInsert(mlir::DistinctAttr)

renxida

/home/azureuser/iree-build/tools/iree-compile: /home/azureuser/miniconda/lib/libtinfo.so.6: no version information available (required by /home/azureuser/iree-build/lib/libIREECompiler.so)
iree-compile: iree/third_party/llvm-project/llvm/include/llvm/Support/Casting.h:566: decltype(auto) llvm::cast(const From &) [To = mlir::DenseElementsAttr, From = mlir::Attribute]: Assertion `isa<To>(Val) && "cast<Ty>() argument of incompatible type!"' failed.
Please report issues to https://github.com/openxla/iree/issues and include the crash backtrace.
Stack dump:
0.	Program arguments: /home/azureuser/iree-build/tools/iree-compile --iree-hal-target-backends=llvm-cpu opt-125M.fp32.onnx.torch.mlir
Stack dump without symbol names (ensure you have llvm-symbolizer in your PATH or set the environment var `LLVM_SYMBOLIZER_PATH` to point to it):
0  libIREECompiler.so 0x00007fed01436997 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) + 39
1  libIREECompiler.so 0x00007fed01434bc0 llvm::sys::RunSignalHandlers() + 80
2  libIREECo

renxida

/home/azureuser/iree-build/tools/iree-compile: /home/azureuser/miniconda/lib/libtinfo.so.6: no version information available (required by /home/azureuser/iree-build/lib/libIREECompiler.so)
Args: /home/azureuser/iree-build/tools/iree-compile --iree-hal-target-backends=llvm-cpu -o output.vmfb stripped-opt-125M.fp32.onnx.torch.mlir --debug 
Load new dialect in Context builtin
ImplicitTypeIDRegistry::lookupOrInsert(mlir::ShapedType)
ImplicitTypeIDRegistry::lookupOrInsert(mlir::MemRefLayoutAttrInterface)
ImplicitTypeIDRegistry::lookupOrInsert(mlir::TypedAttr)
ImplicitTypeIDRegistry::lookupOrInsert(mlir::ElementsAttr)
ImplicitTypeIDRegistry::lookupOrInsert(mlir::DistinctAttr)
ImplicitTypeIDRegistry::lookupOrInsert(mlir::BytecodeOpInterface)
ImplicitTypeIDRegistry::lookupOrInsert(mlir::SymbolOpInterface)

renxida

import onnx
import numpy as np


from onnx import numpy_helper, TensorProto, save_model
from onnx.helper import make_model, make_node, make_graph, make_tensor_value_info
from onnx.checker import check_model

# condition has to be a float tensor
condition = make_tensor_value_info('condition', TensorProto.FLOAT, [1])

renxida

import gc
import sys
import torch
import torch_mlir

batch_size = 1
seq_len = 3
input_size = 5
hidden_size = 5
kernel_size = 3

renxida

Add
AveragePool
BatchNormalization
Cast
Clip
Concat
Constant
ConstantOfShape
Conv

renxida

// torch-mlir/.vscode/settings.json
{
    "files.associations": {
        "*.inc": "cpp",
        "ranges": "cpp",
        "regex": "cpp",
        "functional": "cpp",
        "chrono": "cpp",
        "__functional_03": "cpp",
        "target": "cpp",

renxida

OVERVIEW: MLIR modular optimizer driver

Available Dialects: builtin, chlo, complex, func, linalg, memref, ml_program, scf, sparse_tensor, stablehlo, tensor, tm_tensor, torch, torch_c, tosa, vhlo
USAGE: torch-mlir-opt [options] <input file>

OPTIONS:

Color Options:

  --color                                                  - Use colors in output (default=autodetect)