Common library code snippet pytorch_ based【tips】

torch library

torch.empty(): Returns an uninitialized tensor .

torch.nn.softmax(): Use softmax function .pytorch:nn.Softmax()_ Unadorned melody blog -CSDN Blog _pytorch softmax function Deep learning softmax Function understanding _ Maverick's blog -CSDN Blog _ Deep learning softmax

torch.nn.AvgPool2d(): The average pooling .nn.AvgPool2d() function _ Code farmer's blog of Lanxiang technical school -CSDN Blog _avgpool2d

torch.nn.Linear(): Whole company class .PyTorch Of nn.Linear（） Detailed explanation _ When the wind and snow return at night o The blog of -CSDN Blog _nn.linear Other methods

torch.nn.ConvTransposed(): deconvolution .

torch.nn.utils.clip_grad_norm_(): Gradient cut （Clipping Gradient）：torch.nn.utils.clip_grad_norm_aabb7654321 The blog of -CSDN Blog

torch.nn.Parameter(): Let some variables constantly modify their values in the process of learning to achieve optimization .PyTorch Medium torch.nn.Parameter() Detailed explanation _Adenialzz The blog of -CSDN Blog _nn parameter

.requires_grad=True: Track calculation records .detach()、data、with no_grad()、requires_grad The relationship between _ Unbearable blog -CSDN Blog

torch.tensor(): Depending on the type of data entered , Output corresponding tensor tensor , The data can be list/tuple/array/scalar.torch.Tensor and torch.tensor The difference between _Vic_Hao The blog of -CSDN Blog _torch.t

torch.detach():pytorch Medium .detach() .detach_()_ Xiaoyihong's blog -CSDN Blog _.detach()

torch.index_select():pytorch Function torch.index_select_ Thigh strong blog -CSDN Blog _torch.index_select

torch.mean(): Return mean ,keepdim=True Then keep the original tensor dimension unchanged , Otherwise, it will disappear after calculating the mean value .pytorch in tensor.mean(axis, keepdim) Parameter understanding small experiment _ Pumpkin pie three vegetables blog -CSDN Blog _tensor.mean

torch.range() and torch.arrange():arrange It doesn't contain end,range contain end, Created tensor The types are different .pytorch.range() and pytorch.arange() The difference between _Who is abc The blog of -CSDN Blog _torch.arrange

torch.matmul(): Tensor multiplication torch.matmul() Tensor multiplication _y4ung The blog of -CSDN Blog _torch.matmul

torch.pow(): Power the input tensor every fraction .torch.randn()、torch.mean()、torch.pow()、torch.matmul()_ Only a cautious blog -CSDN Blog _torch.randn()

torch.topk(): Return tensor K The biggest / Small value .torch.topk() Function details _ewahiogj The blog of -CSDN Blog _torch.topk()

torch.nn.functional.pad(): Expand the image matrix or characteristic matrix .F.pad() Function details _ Xiaoyezhu's blog -CSDN Blog _f.pad

torch.mul(): Dot matrix .torch.mul()—— Matrix point multiplication _ Visual sprouting 、 The blog of -CSDN Blog _torch Dot matrix

.item(): take tensor The tensor is converted to python Scalar .Pytorch in .numpy() .item() .cpu() difference _ Ma Pengsen's blog -CSDN Blog

numpy library

numpy.logx()： With x Log base .

np.concatenate(): Yes array Spliced functions .

np.random.permutation(): Generate a scrambled list for a given list .

np.zeros_like(): Output has the same structure as input 0 matrix .numpy And zeros_like()_Nicola-Zhang The blog of -CSDN Blog _numpy zeros_like

.min(),.max(),.mean():numpy Array for maximum and minimum .numpy.max In the parameter axis The problem of the value of _Zpadger The blog of -CSDN Blog _max(axis=0)

np.percentile(): Any percentage digit of the array .

【python】numpy library np.percentile Detailed explanation _brucewong0516 The blog of -CSDN Blog _numpy.percentile

np.clip():np.clip() usage Numpy in clip Use of functions _yctjin The blog of -CSDN Blog _np.clip

np.expand_dims(): Expand the shape of the array .np.expand_dims Xiaobai explained in detail _ watermelon 6 The blog of -CSDN Blog _np.expand_dims()

np.astype(): transformation numpy Array .np.astype()_ The plum blossom 14 The blog of -CSDN Blog _.astype

np.uint8:Opencv numpy in uint8 Type store image _ Programmer hair loss blog -CSDN Blog _np.uint8

np.cov():【[Numpy Learning record 】np.cov Detailed explanation _jeffery0207 The blog of -CSDN Blog _np.cov() Mathematical principle of covariance matrix ,numpy Calculate the covariance matrix (np.cov) Function explanation and source code analysis _Codefmeister The blog of -CSDN Blog _np Calculate covariance

np.astype(): transformation numpy Data type of array .np.astype()_ The plum blossom 14 The blog of -CSDN Blog _.astype

np.unit8:uint8 It is specially used to store various images （ Include RGB, Gray scale image, etc ）, Range 0–255.Opencv numpy in uint8 Type store image _ Programmer hair loss blog -CSDN Blog _np.uint8

torchvision library

transforms.ColorJitter(): Data to enhance , Image brightness can be changed / Contrast / saturation / tonal .PyTorch - transforms.ColorJitter Change the properties of the image ： brightness （brightness）、 Contrast （contrast）、 saturation （saturation） And hue （hue)_TheOldManAndTheSea The blog of -CSDN Blog _transforms.colorjitter

functools library

functools.partial(): Partial function , It is applicable to a large number of calls to the same function with different default parameter values , Make a call platform . Detailed explanation Python Partial functions in （Partial Function）-functools.partial_ Super super genius blog -CSDN Blog

random library

random.uniform(): Generate random real numbers , Including the minimum value .[Python] random.uniform( ) Function tutorial and instance parsing _Harry Wow blog -CSDN Blog _python random.uniform

cv2 library

cv2.pyrDown(): Down sample the original image . First, Gaussian transform , Then discard even lines , Even columns realize down sampling . The first 11 Chapter Image pyramid -- Down sampling cv2.pyrDown() , On the sampling cv2.pyrUp()_Enzo Blog that wants to smash the computer -CSDN Blog _cv2 Down sampling

cv2.pyrUp():pyrDown() Anti operation of .

cv2.subtract(): The two images are subtracted .OpenCV-Python Subtraction of images cv2.subtract Function details and comparison with matrix subtraction _LaoYuanPython The blog of -CSDN Blog _opencv subtract

cv.getStructuringElement(): use 0,1 Building structural elements . Structured elements cv2.getStructuringElement()_Xahoo！ The blog of -CSDN Blog _ Structured elements

cv2.morphologyEx(): Carry out various morphological transformations .opencv-python And cv2.morphologyEx（）_ Yizhi bubble's blog -CSDN Blog _morphologyex python

model operation

model.train(): The model has BN layer ,Dropout add . Yes BN The layer ensures that the average value of each batch of input data is updated , variance .

model.eval(): Guarantee BN The layer directly uses the mean variance obtained from the previous training , Ensure that the mean variance remains unchanged during the test .pytorch Learning notes - In network training ,model.train() model.eval() Use _ Up in the sky ！ The blog of -CSDN Blog name

matplotlib library

plt.clf(): Clear all axes , The window opens .【Matplotlib】 clear axes and figure（plt.cla、plt.clf、plt.close）_ZSYL The blog of -CSDN Blog _matplotlib Empty axes

plt.subplot(): Divide the display interface into grids .matplotlib Medium plt.subplot() Introduction _ The blog of the little famous cute ghost -CSDN Blog _plt.subplot

Built in functions

set(): Is an unordered sequence of non repeating elements . Frequently used {} or set() Function representation .

.index():Python Medium index() Method _ Beimu . The blog of -CSDN Blog _python in index

sample():(a,n)：a Is a sequence ,n For random from a Mid extraction n Number .python sample function _ Dr. Xiang's blog -CSDN Blog _python sample function

.values:Python In the dictionary values function _Insane_Loafer The blog of -CSDN Blog _python in value Function usage

-------------------- Ongoing update --------------------