Convert Pytorch model to onnx or tflite, and the converted model can be visualized by Netron

Last update: Nov 19, 2022

Related tags

Deep Learning model-converter

Overview

Model converter

Convert Pytorch model to onnx or tflite, and the converted model can be visualized by Netron

You can use this project to:

Pytorch -> onnx (float32)
Pytorch -> onnx -> tflite (float32)
Pytorch -> onnx -> tflite (int8)

Requirements

torch2onnx

pytorch
onnx
opencv-python

torch2tflite

tensorflow ~= 2.5
torch == 1.8.1
tensorflow-addons ~= 0.15
opencv-python ~= 4.5.4
onnx ~= 1.10
onnx-tf ~= 1.9
numpy >= 1.19

(opencv-python is optional)

Usage

torch2onnx (float32)

from converter import Torch2onnxConverter

converter = Torch2onnxConverter(model_path, target_shape=(3,224,224))
converter.convert()

torch2tflite (float32)

from converter import Torch2TFLiteConverter

converter = Torch2TFLiteConverter(tmp_path, tflite_model_save_path='model_float32.lite', target_shape=(224,224,3))
converter.convert()

torch2tflite (int8)

from converter import Torch2TFLiteConverter

converter = Torch2TFLiteConverter(tmp_path, tflite_model_save_path='model_int8.lite', target_shape=(224,224,3),
                                    representative_dataset=representative_dataset)
converter.convert()

More details can be found in Torch2onnxConverter and Torch2TfliteConverter __init__ method.

Note that target_shape is different for Pytorch and Tensorflow.

Convert Pytorch model to onnx or tflite, and the converted model can be visualized by Netron

Related tags

Overview

Model converter

Requirements

Usage

Example

Owner

Roxbili

Notepy is a full-featured Notepad Python app

Proto-RL: Reinforcement Learning with Prototypical Representations

Project looking into use of autoencoder for semi-supervised learning and comparing data requirements compared to supervised learning.

Real-CUGAN - Real Cascade U-Nets for Anime Image Super Resolution

Neural Surface Maps

Official code for Next Check-ins Prediction via History and Friendship on Location-Based Social Networks (MDM 2018)

[ICCV 2021 Oral] NerfingMVS: Guided Optimization of Neural Radiance Fields for Indoor Multi-view Stereo

Learning Optical Flow from a Few Matches (CVPR 2021)

x-transformers-paddle 2.x version

A Pythonic library for Nvidia Codec.

FreeSOLO for unsupervised instance segmentation, CVPR 2022

[ICML 2022] The official implementation of Graph Stochastic Attention (GSAT).

A facial recognition doorbell system using a Raspberry Pi

Pytorch implementation of DeepMind's differentiable neural computer paper.

SparseInst: Sparse Instance Activation for Real-Time Instance Segmentation, CVPR 2022

The Generic Manipulation Driver Package - Implements a ROS Interface over the robotics toolbox for Python

EasyMocap is an open-source toolbox for markerless human motion capture from RGB videos.

Saliency - Framework-agnostic implementation for state-of-the-art saliency methods (XRAI, BlurIG, SmoothGrad, and more).

Code for paper "ASAP-Net: Attention and Structure Aware Point Cloud Sequence Segmentation"

Generic U-Net Tensorflow implementation for image segmentation