程永强

ubuntu 16.04 Caffe2 / PyTorch - CUDA / CUDNN 信息1. CUDA / CUDNN 信息yongqiang@famu-sys:~$ pythonPython 2.7.12 (default, Dec 4 2017, 14:50:18) [GCC 5.4.0 20160609] on linux2Type "help", "copyright", "credits" or "license" for more information.>>&g

PyTorch - torch.onnxTable of Contents (目录)PyTorch Tutorialshttps://pytorch.org/tutorials/PyTorch Documentationhttps://pytorch.org/docs/stable/index.htmltorch.onnxhttps://pytorch.org/docs/stable/onnx.htmlGitHub - PyTorchhttps://github.com/pytorch/p

PyTorch model - optimizer - state_dict() - torch.save(config, save_path) - torch.load(load_path)1. state_dict()#!/usr/bin/env python# -*- coding: utf-8 -*-# yongqiang chengfrom __future__ import absolute_importfrom __future__ import divisionfrom __

Keras style model.summary() in PyTorch - 打印 PyTorch 模型结构https://github.com/sksq96/pytorch-summaryhttps://pypi.org/project/torchsummary/Model summary in PyTorch similar to model.summary() in Keras.Keras has a neat API to view the visualization of the mo

PyTorch - directoryPyTorch Tutorialshttps://pytorch.org/tutorials/PyTorch Documentationhttps://pytorch.org/docs/stable/index.htmltorch.onnxhttps://pytorch.org/docs/stable/onnx.htmlGitHub - PyTorchhttps://github.com/pytorch/pytorch1. c10 - Caffe Te

UnavailableInvalidChannel: The channel is not accessible or is invalid.1. UnavailableInvalidChannel: The channel is not accessible or is invalid.(base) yongqiang@famu-sys:~$ conda install pytorch==1.1.0 torchvision==0.3.0 cudatoolkit=9.0Collecting packa

PyTorch 中文文档 - 教程PyTorch 中文文档https://www.pytorchtutorial.com/docs/https://pytorch-cn.readthedocs.io/zh/latest/PyTorch 中文文档 & 教程https://pytorch.apachecn.org/

saving and loading models - 保存和读取模型https://pytorch.org/tutorials/beginner/saving_loading_models.htmlhttps://github.com/pytorch/tutorials/blob/master/beginner_source/saving_loading_models.pyhttps://pytorch.org/tutorials/index.htmlThis document provides

PyTorch softmax - torch.nn - torch.nn.functionalPyTorch documentationhttps://pytorch.org/docs/stable/index.html0. softmax - torch.nn - torch.nn.functionaltorch.nn.functional.softmax (Python function, in torch.nn.functional)定义的是一个函数。torch.nn.functio.

PyTorch torch.Tensor

torch.tensor - v1.5.0PyTorch documentationhttps://pytorch.org/docs/stable/torch.tensor (Python function, in torch)torch.Tensor (Python class, in torch.Tensor)torch.tensor(data, dtype=None, device=None, requires_grad=False, pin_memory=False) -> Te

torch.Tensor.permute - 数据重排重新排列维度，不改变维度。1. torch.Tensor.permutePython method, in torch.Tensor1.1 permute(*dims) -> TensorReturns a view of the original tensor with its dimensions permuted.返回其尺寸已排列的原始张量的视图。1.2 Parameters*dims (int...) - The desir

PyTorch documentationPick a versionhttps://pytorch.org/docs/versions.htmlv1.1.0https://pytorch.org/docs/1.1.0/PyTorch is an optimized tensor library for deep learning using GPUs and CPUs.NotesCommunityPackage Referencetorchvision Reference

Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference (纯整数算术计算)Benoit Jacob, Skirmantas Kligys, Bo Chen, Menglong Zhu, Matthew Tang, Andrew Howard, Hartwig Adam, Dmitry Kalenichenkopost-training quantization，PTQ：训练后

TensorFlow Lite post-training quantization (PTQ，训练后量化)Post-training quantization (PTQ) is a conversion technique that can reduce model size while also improving CPU and hardware accelerator latency, with little degradation in model accuracy. You can quant

TensorFlow Lite 8-bit quantization specification (8-bit 量化规范)The following document outlines the specification for TensorFlow Lite's 8-bit quantization scheme. This is intended to assist hardware developers in providing hardware support for inference with

TensorFlow Lite model optimization (模型优化)Edge devices often have limited memory or computational power. Various optimizations can be applied to models so that they can be run within these constraints. In addition, some optimizations allow the use of speci

TensorFlow v2.8.0 - Android NDK API - space_to_depth1. tf.nn.space_to_depthTensorFlowhttps://www.tensorflow.org/TensorFlow -> API -> TensorFlow Core v2.8.0 -> Pythonhttps://www.tensorflow.org/api_docs/python/tf/nn/space_to_depthSpaceToDepth

TensorFlow - Post-training quantizationTensorFlowhttps://github.com/tensorflow/tensorflowTensorFlow - litehttps://github.com/tensorflow/tensorflow/tree/master/tensorflow/litetensorflow/lite/g3doc/performancehttps://github.com/tensorflow/tensorflow/tr

TensorFlow 0.x - TensorFlow 1.x - TensorFlow 2.x1. TensorFlow 0.x and TensorFlow 1.x静态图，效率高，容易优化，但是调试麻烦。2. TensorFlow 2.x动态图，不易优化，但是调试方便。yongqiang cheng调试方便十分重要。...

面向机器智能的 TensorFlow 实践1.English EditionTensorFlow for Machine Intelligencehttps://bleedingedgepress.com/tensor-flow-for-machine-intelligence/2. Code & Erratabackstopmedia / tensorflowbookhttps://github.com/backstopmedia/tensorflowbook3.C..

Keras Documentation - Keras 文档Keras is a high-level neural networks API, written in Python and capable of running on top of TensorFlow, CNTK, or Theano. It was developed with a focus on enabling fast experimentation. Being able to go from idea to result w

tf.argmax (API r1.3)https://github.com/tensorflow/docs/tree/r1.3/site/en/api_docs/api_docs/python/tfsite/en/api_docs/api_docs/python/tf/argmax.mdargmax( input, axis=None, name=None, dimension=None, output_type=tf.int64)Defined in

tf.matmul (API r1.3) - 矩阵乘法https://github.com/tensorflow/docs/tree/r1.3/site/en/api_docs/api_docs/python/tfmatmul( a, b, transpose_a=False, transpose_b=False, adjoint_a=False, adjoint_b=False, a_is_sparse=False, b_is_spars

tf.concathttps://github.com/tensorflow/docs/tree/r1.4/site/en/api_docs/api_docs/python/tfsite/en/api_docs/api_docs/python/tf/concat.mdconcat( values, axis, name='concat')Defined in tensorflow/python/ops/array_ops.py.See the guide: Tensor

tf.split (API r1.3)https://github.com/tensorflow/docs/blob/r1.3/site/en/api_docs/api_docs/python/tf/split.mdsite/en/api_docs/api_docs/python/tf/split.mdsplit( value, num_or_size_splits, axis=0, num=None, name='split')Defined in te

tf.reshape (API r1.3)https://github.com/tensorflow/docs/tree/r1.3/site/en/api_docs/api_docs/python/tfsite/en/api_docs/api_docs/python/tf/reshape.mdreshape( tensor, shape, name=None)Defined in tensorflow/python/ops/gen_array_ops.py.See th

tf.to_int32https://github.com/tensorflow/docs/tree/r1.4/site/en/api_docs/api_docs/python/tfsite/en/api_docs/api_docs/python/tf/to_int32.mdto_int32( x, name='ToInt32')Defined in tensorflow/python/ops/math_ops.py.See the guide: Tensor Transfor

tf.equal (API r1.3)https://github.com/tensorflow/docs/tree/r1.3/site/en/api_docs/api_docs/python/tfsite/en/api_docs/api_docs/python/tf/equal.mdequal( x, y, name=None)Defined in tensorflow/python/ops/gen_math_ops.py.See the guide: Control

tf.transpose (API r1.3)https://github.com/tensorflow/docs/tree/r1.3/site/en/api_docs/api_docs/python/tfsite/en/api_docs/api_docs/python/tf/transpose.mdtranspose( a, perm=None, name='transpose')Defined in tensorflow/python/ops/array_ops.py

Conda - Python 3.5 - Tensorflow 1.4.11. conda 创建 python=3.x 虚拟环境 $ENVIRONMENT_NAMEconda create --name $ENVIRONMENT_NAME python=3.x - conda 创建 python=3.x 虚拟环境 $ENVIRONMENT_NAME(base) yongqiang@yongqiang:~$ conda create --name tf_cpu_1.4.1 python=3.6Coll

从源代码构建 TensorFlow 安装配置 build通过运行 TensorFlow 源代码树根目录下的 ./configure 配置系统 build。此脚本会提示您指定 TensorFlow 依赖项的位置，并要求指定其他构建配置选项。./configure以下 ./configure 脚本展示了 ./configure 的示例运行会话，查看示例配置会话。 You have bazel 0.15.0 installed. Please specify the location o

TensorFlow 正则化 (regularization) 与过拟合1. L1L1L1 L2L2L2 正则化与过拟合在使用梯度下降优化神经网络时，被优化的函数就是神经网络的损失函数。这个损失函数刻画了在训练数据集上预测结果和真实结果之间的差距。然而在真实的应用中，我们想要的并不是让模型尽量模拟训练数据的行为，而是希望通过训练出来的模型对未知的数据给出判断。模型在训练数据上的表现并不一定代表了它在未知数据上的表现。过拟合问题就是可以导致这个差距的一个很重要因素。过拟合指的是当一个模型过为复杂之后，它

TensorFlow-Slimhttps://github.com/tensorflow/tensorflow/tree/r1.4/tensorflow/contrib/slimdecay [dɪ'keɪ]：v. 衰退，衰减，衰败，腐烂，腐朽 n. 衰退，衰减，衰败，腐烂，腐朽parametric [pærə'metrɪk]：adj. 参数的，参量的scope [skəʊp]：n. 范围，余地，视野，眼界，导弹射程 vt. 审视slim [slɪm]：adj. 苗条的，修长的，微小的，差的 vt.

tensorflow 安装 - tensorflow cpu - gpu - PyPIhttps://pypi.org/project/tensorflow/https://pypi.org/project/tensorflow-gpu/TensorFlow is an open source machine learning framework for everyone.1. Project descriptionTensorFlow is an open source software lib

TensorFlow 对应的 CUDA 和 cuDNN 版本配置https://www.tensorflow.org/install/source要了解可用于旧版 TensorFlow 的 CUDA 和 cuDNN 版本，请参阅经过测试的编译配置。1. Linux1.1 CPU1.2 GPU2. macOS2.1 CPU2.2 GPU

TensorFlow-ExamplesTensorFlow Tutorial and Examples for Beginners (TF v1 & v2)https://github.com/aymericdamien/TensorFlow-Exampleshttps://github.com/aymericdamien/TensorFlow-Examples/tree/0.11

ERROR: markdown 3.1.1 has requirement setuptools>=36, but you’ll have setuptools 20.7.0 which is incompatible.1. ERROR: markdown 3.1.1 has requirement setuptools>=36, but you’ll have setuptools 20.7.0 which is incompatible.yongqiang@famu-sys:~$ sud

Command “python setup.py egg_info” failed with error code 1 in /tmp/pip-install-3wDHN2/html5lib/1. sudo pip install tensorflow-gpu==1.4.0 - tensorflow_gpu-1.4.0-cp35-cp35m-manylinux1_x86_64.whl注意使用 sudo pip install tensorflow-gpu==1.4.0 安装 tensorflow_gpu

tf.app.flags解析命令行参数。1. /usr/local/lib/python2.7/dist-packages/tensorflow/python/platform/flags.py"""Implementation of the flags interface."""from __future__ import absolute_importfrom __future__ import divisionfrom __future__ import print_functioni