"Use a toy dataset to train a classification model" is a simplest deep learning practice.
Today I want to record how to use MNIST A HANDWRITTEN DIGIT RECOGNITION dataset to build a simple classifier in PyTorch.
Read More »[PyTorch] Tutorial(4) Train a model to classify MNIST dataset
The so-called Neural Network is the model architecture we want to build for deep learning. In official PyTorch document, the first sentence clearly states:
You can use torch.nn to build a neural network.
I believe everyone who is familiar with deep learning knows the importance of automatic derivation. In here, I will make a simple record.
The neural network operation of the deep learning model includes forward propagation and backward propagation.
Read More »[PyTorch] Tutorial(2) Automatic derivative 
I often got this error message when I learning about machine learning by Python, recently.
'Object arrays cannot be loaded when allow_pickle=False'The reason for this message is actually very clearly. In fact, it's just that "allow_pickle" is adjusted to False after Numpy package is upgraded.
And it caused us to be hindered when reading some pickle-encapsulated training data.
Read More »[Solved] 'Object arrays cannot be loaded when allow_pickle=False' 
If you are familiar with the deep learning and its environment, you may often hear "CUDA". Before we starting to train our model, maybe we need to install GPU driver, CUDA and cuDNN toolkits. They can make our training have the best effect.
Read More »Easy way to remove CUDA I have been studying about pointer-generator network lately, and I found a package "pkbar"can let us to use keras style progress bar in our tasks.
In fact, it can be used not only in the code of PyTorch framework but also in any loop.
Read More »[PyTorch] A progress bar using Keras style: pkbar
It needs to be stated in advance that the teaching of this article is learning from: https://towardsdatascience.com/object-detection-with-10-lines-of-code-d6cb4d86f606. You can also go directly to that website to learn.
Read More »[Machine Learning][Python] The simplest object recognition package: ImageAI Today when I reading the document of the "Transformers" package which Hugging Face developed, I suddenly discovered the model I build, its state is "eval()", not "train()". It seems that as long as we use "from_pretrained()" method is the default state "eval()".
Read More »[PyTorch] How to check the model state is "train()" or "eval()" This is very common error message in PyTorch.
RuntimeError: bool value of Tensor with more than one value is ambiguousWhen I using PyTorch to train a model, I often use GPU_A to train the model, save model. But if I load the model I saved to test some new data, I always put the new data in a different GPU, we called it GPU_B.
Read More »[PyTorch] How to check which GPU device our data used