Simple ONNX operation generator. Simple Operation Generator for ONNX.

Overview

sog4onnx

Simple ONNX operation generator. Simple Operation Generator for ONNX.

https://github.com/PINTO0309/simple-onnx-processing-tools

Downloads GitHub PyPI CodeQL

Key concept

  • Variable, Constant, Operation and Attribute can be generated externally.
  • Allow Opset to be specified externally.
  • No check for consistency of Operations within the tool, as new OPs are added frequently and the definitions of existing OPs change with each new version of ONNX's Opset.
  • Only one OP can be defined at a time, and the goal is to generate free ONNX graphs using a combination of snc4onnx, sne4onnx, snd4onnx and scs4onnx.
  • List of parameters that can be specified: https://github.com/onnx/onnx/blob/main/docs/Operators.md

1. Setup

1-1. HostPC

### option
$ echo export PATH="~/.local/bin:$PATH" >> ~/.bashrc \
&& source ~/.bashrc

### run
$ pip install -U onnx \
&& python3 -m pip install -U onnx_graphsurgeon --index-url https://pypi.ngc.nvidia.com \
&& pip install -U sog4onnx

1-2. Docker

### docker pull
$ docker pull pinto0309/sog4onnx:latest

### docker build
$ docker build -t pinto0309/sog4onnx:latest .

### docker run
$ docker run --rm -it -v `pwd`:/workdir pinto0309/sog4onnx:latest
$ cd /workdir

2. CLI Usage

$ sog4onnx -h

usage: sog4onnx [-h]
  --op_type OP_TYPE
  --opset OPSET
  --op_name OP_NAME
  [--input_variables NAME TYPE VALUE]
  [--output_variables NAME TYPE VALUE]
  [--attributes NAME DTYPE VALUE]
  [--output_onnx_file_path OUTPUT_ONNX_FILE_PATH]
  [--non_verbose]

optional arguments:
  -h, --help
        show this help message and exit

  --op_type OP_TYPE
        ONNX OP type.
        https://github.com/onnx/onnx/blob/main/docs/Operators.md

  --opset OPSET
        ONNX opset number.

  --op_name OP_NAME
        OP name.

  --input_variables NAME DTYPE VALUE
        input_variables can be specified multiple times.
        --input_variables variable_name numpy.dtype shape
        https://github.com/onnx/onnx/blob/main/docs/Operators.md

        e.g.
        --input_variables i1 float32 [1,3,5,5] \
        --input_variables i2 int32 [1] \
        --input_variables i3 float64 [1,3,224,224]

  --output_variables NAME DTYPE VALUE
        output_variables can be specified multiple times.
        --output_variables variable_name numpy.dtype shape
        https://github.com/onnx/onnx/blob/main/docs/Operators.md

        e.g.
        --output_variables o1 float32 [1,3,5,5] \
        --output_variables o2 int32 [1] \
        --output_variables o3 float64 [1,3,224,224]

  --attributes NAME DTYPE VALUE
        attributes can be specified multiple times.
        dtype is one of "float32" or "float64" or "int32" or "int64" or "str".
        --attributes name dtype value
        https://github.com/onnx/onnx/blob/main/docs/Operators.md

        e.g.
        --attributes alpha float32 1.0 \
        --attributes beta float32 1.0 \
        --attributes transA int32 0 \
        --attributes transB int32 0

  --output_onnx_file_path OUTPUT_ONNX_FILE_PATH
        Output onnx file path.
        If not specified, a file with the OP type name is generated.

        e.g. op_type="Gemm" -> Gemm.onnx

  --non_verbose
        Do not show all information logs. Only error logs are displayed.

3. In-script Usage

$ python
>>> from sog4onnx import generate
>>> help(generate)
Help on function generate in module sog4onnx.onnx_operation_generator:

generate(
  op_type: str,
  opset: int,
  op_name: str,
  input_variables: dict,
  output_variables: dict,
  attributes: Union[dict, NoneType] = None,
  output_onnx_file_path: Union[str, NoneType] = '',
  non_verbose: Union[bool, NoneType] = False
) -> onnx.onnx_ml_pb2.ModelProto

    Parameters
    ----------
    op_type: str
        ONNX op type.
        See below for the types of OPs that can be specified.
        https://github.com/onnx/onnx/blob/main/docs/Operators.md

        e.g. "Add", "Div", "Gemm", ...

    opset: int
        ONNX opset number.

        e.g. 11

    op_name: str
        OP name.

    input_variables: Optional[dict]
        Specify input variables for the OP to be generated.
        See below for the variables that can be specified.
        https://github.com/onnx/onnx/blob/main/docs/Operators.md
        {"input_var_name1": [numpy.dtype, shape], "input_var_name2": [dtype, shape], ...}

        e.g.
        input_variables = {
          "name1": [np.float32, [1,224,224,3]],
          "name2": [np.bool_, [0]],
          ...
        }

    output_variables: Optional[dict]
        Specify output variables for the OP to be generated.
        See below for the variables that can be specified.
        https://github.com/onnx/onnx/blob/main/docs/Operators.md
        {"output_var_name1": [numpy.dtype, shape], "output_var_name2": [dtype, shape], ...}

        e.g.
        output_variables = {
          "name1": [np.float32, [1,224,224,3]],
          "name2": [np.bool_, [0]],
          ...
        }

    attributes: Optional[dict]
        Specify output attributes for the OP to be generated.
        See below for the attributes that can be specified.
        When specifying Tensor format values, specify an array converted to np.ndarray.
        https://github.com/onnx/onnx/blob/main/docs/Operators.md
        {"attr_name1": value1, "attr_name2": value2, "attr_name3": value3, ...}

        e.g.
        attributes = {
          "alpha": 1.0,
          "beta": 1.0,
          "transA": 0,
          "transB": 0
        }
        Default: None

    output_onnx_file_path: Optional[str]
        Output of onnx file path.
        If not specified, no .onnx file is output.
        Default: ''

    non_verbose: Optional[bool]
        Do not show all information logs. Only error logs are displayed.
        Default: False

    Returns
    -------
    single_op_graph: onnx.ModelProto
        Single op onnx ModelProto

4. CLI Execution

$ sog4onnx \
--op_type Gemm \
--opset 1 \
--op_name gemm_custom1 \
--input_variables i1 float32 [1,2,3] \
--input_variables i2 float32 [1,1] \
--input_variables i3 int32 [0] \
--output_variables o1 float32 [1,2,3] \
--attributes alpha float32 1.0 \
--attributes beta float32 1.0 \
--attributes transA int32 0 \
--attributes transB int32 0

5. In-script Execution

import numpy as np
from sog4onnx import generate

single_op_graph = generate(
    op_type = 'Gemm',
    opset = 1,
    op_name = "gemm_custom1",
    input_variables = {
      "i1": [np.float32, [1,2,3]],
      "i2": [np.float32, [1,1]],
      "i3": [np.int32, [0]],
    },
    output_variables = {
      "o1": [np.float32, [1,2,3]],
    },
    attributes = {
      "alpha": 1.0,
      "beta": 1.0,
      "broadcast": 0,
      "transA": 0,
      "transB": 0,
    },
    non_verbose = True,
)

6. Sample

6-1. opset=1, Gemm

$ sog4onnx \
--op_type Gemm \
--opset 1 \
--op_name gemm_custom1 \
--input_variables i1 float32 [1,2,3] \
--input_variables i2 float32 [1,1] \
--input_variables i3 int32 [0] \
--output_variables o1 float32 [1,2,3] \
--attributes alpha float32 1.0 \
--attributes beta float32 1.0 \
--attributes transA int32 0 \
--attributes transB int32 0
--non_verbose

image image

6-2. opset=11, Add

$ sog4onnx \
--op_type Add \
--opset 11 \
--op_name add_custom1 \
--input_variables i1 float32 [1,2,3] \
--input_variables i2 float32 [1,2,3] \
--output_variables o1 float32 [1,2,3] \
--non_verbose

image image

6-3. opset=11, NonMaxSuppression

$ sog4onnx \
--op_type NonMaxSuppression \
--opset 11 \
--op_name nms_custom1 \
--input_variables boxes float32 [1,6,4] \
--input_variables scores float32 [1,1,6] \
--input_variables max_output_boxes_per_class int64 [1] \
--input_variables iou_threshold float32 [1] \
--input_variables score_threshold float32 [1] \
--output_variables selected_indices int64 [3,3] \
--attributes center_point_box int64 1

image image

6-4. opset=11, Constant

$ sog4onnx \
--op_type Constant \
--opset 11 \
--op_name const_custom1 \
--output_variables boxes float32 [1,6,4] \
--attributes value float32 \
[[\
[0.5,0.5,1.0,1.0],\
[0.5,0.6,1.0,1.0],\
[0.5,0.4,1.0,1.0],\
[0.5,10.5,1.0,1.0],\
[0.5,10.6,1.0,1.0],\
[0.5,100.5,1.0,1.0]\
]]

image

7. Reference

  1. https://github.com/onnx/onnx/blob/main/docs/Operators.md
  2. https://docs.nvidia.com/deeplearning/tensorrt/onnx-graphsurgeon/docs/index.html
  3. https://github.com/NVIDIA/TensorRT/tree/main/tools/onnx-graphsurgeon
  4. https://github.com/PINTO0309/sne4onnx
  5. https://github.com/PINTO0309/snd4onnx
  6. https://github.com/PINTO0309/snc4onnx
  7. https://github.com/PINTO0309/scs4onnx
  8. https://github.com/PINTO0309/PINTO_model_zoo

8. Issues

https://github.com/PINTO0309/simple-onnx-processing-tools/issues

You might also like...
Multiple types of NN model optimization environments. It is possible to directly access the host PC GUI and the camera to verify the operation. Intel iHD GPU (iGPU) support. NVIDIA GPU (dGPU) support.
Multiple types of NN model optimization environments. It is possible to directly access the host PC GUI and the camera to verify the operation. Intel iHD GPU (iGPU) support. NVIDIA GPU (dGPU) support.

mtomo Multiple types of NN model optimization environments. It is possible to directly access the host PC GUI and the camera to verify the operation.

Complete system for facial identity system. Include one-shot model, database operation, features visualization, monitoring

Complete system for facial identity system. Include one-shot model, database operation, features visualization, monitoring

Accelerated SMPL operation, commonly used in generate 3D human mesh, STAR included.

SMPL2 An enchanced and accelerated SMPL operation which commonly used in 3D human mesh generation. It takes a poses, shapes, cam_trans as inputs, outp

Liecasadi - liecasadi implements Lie groups operation written in CasADi

liecasadi liecasadi implements Lie groups operation written in CasADi, mainly di

A code generator from ONNX to PyTorch code

onnx-pytorch Generating pytorch code from ONNX. Currently support onnx==1.9.0 and torch==1.8.1. Installation From PyPI pip install onnx-pytorch From

Simple node deletion tool for onnx.
Simple node deletion tool for onnx.

snd4onnx Simple node deletion tool for onnx. I only test very miscellaneous and limited patterns as a hobby. There are probably a large number of bugs

MMdnn is a set of tools to help users inter-operate among different deep learning frameworks. E.g. model conversion and visualization. Convert models between Caffe, Keras, MXNet, Tensorflow, CNTK, PyTorch Onnx and CoreML.
MMdnn is a set of tools to help users inter-operate among different deep learning frameworks. E.g. model conversion and visualization. Convert models between Caffe, Keras, MXNet, Tensorflow, CNTK, PyTorch Onnx and CoreML.

MMdnn MMdnn is a comprehensive and cross-framework tool to convert, visualize and diagnose deep learning (DL) models. The "MM" stands for model manage

PyTorch ,ONNX and TensorRT implementation of YOLOv4
PyTorch ,ONNX and TensorRT implementation of YOLOv4

PyTorch ,ONNX and TensorRT implementation of YOLOv4

YOLOv5 in PyTorch > ONNX > CoreML > TFLite
YOLOv5 in PyTorch ONNX CoreML TFLite

This repository represents Ultralytics open-source research into future object detection methods, and incorporates lessons learned and best practices evolved over thousands of hours of training and evolution on anonymized client datasets. All code and models are under active development, and are subject to modification or deletion without notice.

Comments
  • Small fixes to README

    Small fixes to README

    Thank you for the tool. There are small fixes needed in the README: the attributes of one example missing the type, and the numpy import in another one.

    Otherwise, it works perfectly.

    opened by ibaiGorordo 1
Releases(1.0.15)
  • 1.0.15(Nov 20, 2022)

    • Fixed a bug where Constant and ConstantOfShape opsets were not set

    Full Changelog: https://github.com/PINTO0309/sog4onnx/compare/1.0.14...1.0.15

    Source code(tar.gz)
    Source code(zip)
  • 1.0.14(Sep 8, 2022)

    • Add short form parameter
      $ sog4onnx -h
      
      usage: sog4onnx [-h]
        --ot OP_TYPE
        --os OPSET
        --on OP_NAME
        [-iv NAME TYPE VALUE]
        [-ov NAME TYPE VALUE]
        [-a NAME DTYPE VALUE]
        [-of OUTPUT_ONNX_FILE_PATH]
        [-n]
      
      optional arguments:
        -h, --help
          show this help message and exit
      
        -ot OP_TYPE, --op_type OP_TYPE
          ONNX OP type.
          https://github.com/onnx/onnx/blob/main/docs/Operators.md
      
        -os OPSET, --opset OPSET
          ONNX opset number.
      
        -on OP_NAME, --op_name OP_NAME
          OP name.
      
        -iv INPUT_VARIABLES INPUT_VARIABLES INPUT_VARIABLES, --input_variables INPUT_VARIABLES INPUT_VARIABLES INPUT_VARIABLES
          input_variables can be specified multiple times.
          --input_variables variable_name numpy.dtype shape
          https://github.com/onnx/onnx/blob/main/docs/Operators.md
      
          e.g.
          --input_variables i1 float32 [1,3,5,5] \
          --input_variables i2 int32 [1] \
          --input_variables i3 float64 [1,3,224,224]
      
        -ov OUTPUT_VARIABLES OUTPUT_VARIABLES OUTPUT_VARIABLES, --output_variables OUTPUT_VARIABLES OUTPUT_VARIABLES OUTPUT_VARIABLES
          output_variables can be specified multiple times.
          --output_variables variable_name numpy.dtype shape
          https://github.com/onnx/onnx/blob/main/docs/Operators.md
      
          e.g.
          --output_variables o1 float32 [1,3,5,5] \
          --output_variables o2 int32 [1] \
          --output_variables o3 float64 [1,3,224,224]
      
        -a ATTRIBUTES ATTRIBUTES ATTRIBUTES, --attributes ATTRIBUTES ATTRIBUTES ATTRIBUTES
          attributes can be specified multiple times.
          dtype is one of "float32" or "float64" or "int32" or "int64" or "str".
          --attributes name dtype value
          https://github.com/onnx/onnx/blob/main/docs/Operators.md
      
          e.g.
          --attributes alpha float32 1.0 \
          --attributes beta float32 1.0 \
          --attributes transA int32 0 \
          --attributes transB int32 0
      
        -of OUTPUT_ONNX_FILE_PATH, --output_onnx_file_path OUTPUT_ONNX_FILE_PATH
          Output onnx file path.
          If not specified, a file with the OP type name is generated.
      
          e.g. op_type="Gemm" -> Gemm.onnx
      
        -n, --non_verbose
          Do not show all information logs. Only error logs are displayed.
      
    Source code(tar.gz)
    Source code(zip)
  • 1.0.13(Jun 10, 2022)

  • 1.0.12(Jun 7, 2022)

  • 1.0.11(May 25, 2022)

  • 1.0.10(May 15, 2022)

  • 1.0.9(Apr 26, 2022)

    • Added op_name as an input parameter, allowing OPs to be named.
      • CLI
        sog4onnx [-h]
          --op_type OP_TYPE
          --opset OPSET
          --op_name OP_NAME
          [--input_variables NAME TYPE VALUE]
          [--output_variables NAME TYPE VALUE]
          [--attributes NAME DTYPE VALUE]
          [--output_onnx_file_path OUTPUT_ONNX_FILE_PATH]
          [--non_verbose]
        
      • In-script
        generate(
          op_type: str,
          opset: int,
          op_name: str,
          input_variables: dict,
          output_variables: dict,
          attributes: Union[dict, NoneType] = None,
          output_onnx_file_path: Union[str, NoneType] = '',
          non_verbose: Union[bool, NoneType] = False
        ) -> onnx.onnx_ml_pb2.ModelProto
        
    Source code(tar.gz)
    Source code(zip)
  • 1.0.8(Apr 15, 2022)

  • 1.0.7(Apr 14, 2022)

  • 1.0.6(Apr 14, 2022)

  • 1.0.5(Apr 13, 2022)

  • 1.0.4(Apr 13, 2022)

  • 1.0.3(Apr 12, 2022)

  • 1.0.2(Apr 12, 2022)

  • 1.0.1(Apr 12, 2022)

  • 1.0.0(Apr 12, 2022)

  • 0.0.2(Apr 12, 2022)

  • 0.0.1(Apr 12, 2022)

Owner
Katsuya Hyodo
Hobby programmer. Intel Software Innovator Program member.
Katsuya Hyodo
Simple Dynamic Batching Inference

Simple Dynamic Batching Inference 解决了什么问题? 众所周知,Batch对于GPU上深度学习模型的运行效率影响很大。。。 是在Inference时。搜索、推荐等场景自带比较大的batch,问题不大。但更多场景面临的往往是稀碎的请求(比如图片服务里一次一张图)。 如果

116 Jan 01, 2023
The code of “Similarity Reasoning and Filtration for Image-Text Matching” [AAAI2021]

SGRAF PyTorch implementation for AAAI2021 paper of “Similarity Reasoning and Filtration for Image-Text Matching”. It is built on top of the SCAN and C

Ronnie_IIAU 149 Dec 22, 2022
NeuralTalk is a Python+numpy project for learning Multimodal Recurrent Neural Networks that describe images with sentences.

#NeuralTalk Warning: Deprecated. Hi there, this code is now quite old and inefficient, and now deprecated. I am leaving it on Github for educational p

Andrej 5.3k Jan 07, 2023
HistoSeg : Quick attention with multi-loss function for multi-structure segmentation in digital histology images

HistoSeg : Quick attention with multi-loss function for multi-structure segmentation in digital histology images Histological Image Segmentation This

Saad Wazir 11 Dec 16, 2022
RefineMask (CVPR 2021)

RefineMask: Towards High-Quality Instance Segmentation with Fine-Grained Features (CVPR 2021) This repo is the official implementation of RefineMask:

Gang Zhang 191 Jan 07, 2023
Pytorch implementation for RelTransformer

RelTransformer Our Architecture This is a Pytorch implementation for RelTransformer The implementation for Evaluating on VG200 can be found here Requi

Vision CAIR Research Group, KAUST 21 Nov 22, 2022
Whisper is a file-based time-series database format for Graphite.

Whisper Overview Whisper is one of three components within the Graphite project: Graphite-Web, a Django-based web application that renders graphs and

Graphite Project 1.2k Dec 25, 2022
A pre-trained language model for social media text in Spanish

RoBERTuito A pre-trained language model for social media text in Spanish READ THE FULL PAPER Github Repository RoBERTuito is a pre-trained language mo

25 Dec 29, 2022
(NeurIPS 2020) Wasserstein Distances for Stereo Disparity Estimation

Wasserstein Distances for Stereo Disparity Estimation Accepted in NeurIPS 2020 as Spotlight. [Project Page] Wasserstein Distances for Stereo Disparity

Divyansh Garg 92 Dec 12, 2022
Official PyTorch Implementation of Convolutional Hough Matching Networks, CVPR 2021 (oral)

Convolutional Hough Matching Networks This is the implementation of the paper "Convolutional Hough Matching Network" by J. Min and M. Cho. Implemented

Juhong Min 70 Nov 22, 2022
Code repository for "Reducing Underflow in Mixed Precision Training by Gradient Scaling" presented at IJCAI '20

Reducing Underflow in Mixed Precision Training by Gradient Scaling This project implements the gradient scaling method to improve the performance of m

Ruizhe Zhao 5 Apr 14, 2022
NCVX (NonConVeX): A User-Friendly and Scalable Package for Nonconvex Optimization in Machine Learning.

NCVX NCVX: A User-Friendly and Scalable Package for Nonconvex Optimization in Machine Learning. Please check https://ncvx.org for detailed instruction

SUN Group @ UMN 28 Aug 03, 2022
Matthew Colbrook 1 Apr 08, 2022
[ICCV 2021] Deep Hough Voting for Robust Global Registration

Deep Hough Voting for Robust Global Registration, ICCV, 2021 Project Page | Paper | Video Deep Hough Voting for Robust Global Registration Junha Lee1,

57 Nov 28, 2022
Simple API for UCI Machine Learning Dataset Repository (search, download, analyze)

A simple API for working with University of California, Irvine (UCI) Machine Learning (ML) repository Table of Contents Introduction About Page of the

Tirthajyoti Sarkar 223 Dec 05, 2022
Enabling Lightweight Fine-tuning for Pre-trained Language Model Compression based on Matrix Product Operators

Enabling Lightweight Fine-tuning for Pre-trained Language Model Compression based on Matrix Product Operators This is our Pytorch implementation for t

RUCAIBox 12 Jul 22, 2022
Python/Rust implementations and notes from Proofs Arguments and Zero Knowledge

What is this? This is where I'll be collecting resources related to the Study Group on Dr. Justin Thaler's Proofs Arguments And Zero Knowledge Book. T

Thor 66 Jan 04, 2023
Finite-temperature variational Monte Carlo calculation of uniform electron gas using neural canonical transformation.

CoulombGas This code implements the neural canonical transformation approach to the thermodynamic properties of uniform electron gas. Building on JAX,

FermiFlow 9 Mar 03, 2022
RoboDesk A Multi-Task Reinforcement Learning Benchmark

RoboDesk A Multi-Task Reinforcement Learning Benchmark If you find this open source release useful, please reference in your paper: @misc{kannan2021ro

Google Research 66 Oct 07, 2022
Code release for The Devil is in the Channels: Mutual-Channel Loss for Fine-Grained Image Classification (TIP 2020)

The Devil is in the Channels: Mutual-Channel Loss for Fine-Grained Image Classification Code release for The Devil is in the Channels: Mutual-Channel

PRIS-CV: Computer Vision Group 230 Dec 31, 2022