onnx-mlir/doc/ImportONNXDefs.md

27 lines
2.0 KiB
Markdown
Raw Normal View History

# Import ONNX specifications into ONNX MLIR
2020-01-30 02:45:48 +08:00
The specifications of ONNX are defined under onnx/defs directory in ONNX projects.
There is a python script onnx/defs/gen_doc.py that automatically generate documents about operations in ONNX (docs/Operations.md).
ONNX MLIR modified this script to import ONNX specifications into ONNX MLIR. There are two files generated for ONNX MLIR with the modified gen_doc.py:
2020-01-30 02:45:48 +08:00
1. src/dialect/onnx/onnxop.inc: Operation defintion for MLIR tablegen. Will be included in src/dialect/onnx/onnx.td
2. src/builder/op_build_table.inc: c code for ONNX MLIR frontend to import operation nodes from ONNX model. Will be included in src/builder/frontend_dialect_transformer.cpp
2020-01-30 02:45:48 +08:00
## How to use the script
1. Get ONNX. You can use onnx-mlir/third_party/onnx
2. In your ONNX directory, copy the script docs/gen_doc.py in your ONNX MLIR to onnx/defs in ONNX
2020-01-30 02:45:48 +08:00
3. Run the script: python onnx/defs/gen_doc.py
4. Two files, onnxop.inc and op_buid_table.inc should be generated in current directory
5. copy the two file into your ONNX MLIR: cp onnxop.inc your_onnx-mlir/src/dialect/onnx/onnxop.inc; cp op_build_table.inc your_onnx-mlir/src/builder
6. go to your ONNX MLIR and build
2020-01-30 02:45:48 +08:00
## Consistency
The Operators.md generated by gen_doc.py is copied into doc. Please refer to this specification, not the one in onnx github, to make sure operators are consistent in version with onnxop.inc.
2020-01-30 02:45:48 +08:00
## Customization
In addition to following the ONNX specification, the modified gen_doc.py provides some mechanism for you to customize the output.
2020-01-30 02:45:48 +08:00
Several tables are defined at the beginning of the script:
1. special_attr_defaults: gives attribute special default value.
2. special_op_handler: creates special import function in frontend_dialect_transformer.cpp. Currently special handler is used for operations with oprational arguments
3. ShapeInferenceList: list of operations which has shape inference defined
4. CanonicalList : list of operations which has canonical form
5. manual_code_in_op_def: provides a way to specify any code for an operation in its tablegen