PyTorch深度学习实践（刘二大人）---第四讲反向传播

最新推荐文章于 2024-07-31 14:30:10 发布

小于不是小鱼呀

最新推荐文章于 2024-07-31 14:30:10 发布

阅读量77

点赞数 2

分类专栏： PyTorch 深度学习实践学习笔记文章标签：深度学习 pytorch python

本文链接：https://blog.csdn.net/lyhxty/article/details/130285126

版权

PyTorch 深度学习实践学习笔记专栏收录该内容

6 篇文章 0 订阅

订阅专栏

知识点

在Pytorch进行,tensor（标量、向量、矩阵、高维数据），保存data、grad。

矩阵《The Matrix Cookbook》有很多公式，比如求导等，可以直接查找。

非线性变化函数 Nonlinear function: （后面讲）

求导（链式求导）： $\frac{\partial f(g(x))}{\partial x} = \frac{\partial f(g(x))}{\partial g(x)}\cdot \frac{\partial g(x)}{\partial x}$

代码

import torch

x_data = [1.0, 2.0, 3.0]
y_data = [2.0, 4.0, 6.0]

w = torch.Tensor([1.0])  # 初始化w
w.requires_grad = True  # 是需要计算梯度的


def forward(x):
    return x * w  # w是tensor


def loss(x, y):  # 损失函数 loss=(y'-y)^2=(w*x-y)^2
    y_pred = forward(x)
    return (y_pred - y) ** 2


print('Predict (before training', 4, forward(4).item())

for epoch in range(100):
    sum = 0
    for x, y in zip(x_data, y_data):
        l = loss(x, y)  # 前馈
        l.backward()  # 反向传播
        print('\tgrad:', x, y, w.grad.item())  # item取出数值，标量
        w.data = w.data - 0.01 * w.grad.data  # w.grad梯度，是tensor，必须要取data进行计算，不可张量直接进行计算
        w.grad.data.zero_()  # 清零

        sum += l.item()  # 会释放计算图

    print('Progress:', epoch, l.item())
print("Predict (after training)", 4, forward(4).item())

作业

作业1

作业2

作业3

公式

代码

import torch


# y = x^2+2x+1
x_data = [1.0, 2.0, 3.0]
y_data = [4.0, 9.0, 16.0]

w1 = torch.Tensor([1.0])  # 初始化w
w2 = torch.Tensor([1.0])
b = torch.Tensor([1.0])
w1.requires_grad = True  # 是需要计算梯度的
w2.requires_grad = True
b.requires_grad = True


def forward(x):
    return w1 * x**2 + w2 * x + b  # w是tensor


def loss(x, y):  # 损失函数 loss=(y'-y)^2=(w*x-y)^2
    y_pred = forward(x)
    return (y_pred - y) ** 2


print('Predict (before training', 4, forward(4).item())

for epoch in range(100):
    for x, y in zip(x_data, y_data):
        l = loss(x, y)  # 前馈
        l.backward()  # 反向传播
        print('\tgrad:', x, y, w1.grad.item(), w2.grad.item(), b.grad.item())  # item取出数值，标量
        w1.data = w1.data - 0.01 * w1.grad.data
        w2.data = w2.data - 0.01 * w2.grad.data
        b.data = b.data - 0.01 * b.grad.data  # w.grad梯度，是tensor，必须要取data进行计算，不可张量直接进行计算

        w1.grad.data.zero_()  # 清零
        w2.grad.data.zero_()
        b.grad.data.zero_()

    print('Progress:', epoch, l.item())
print("Predict (after training)", 4, forward(4).item())

output

D:\Conda_data\envs\pytorch\python.exe D:/PycharmProject/Pytorch学习笔记-刘二大人/4.反向传播-作业.py
Predict (before training 4 21.0
   grad: 1.0 4.0 -2.0 -2.0 -2.0
   grad: 2.0 9.0 -14.880001068115234 -7.440000534057617 -3.7200002670288086
   grad: 3.0 16.0 -20.527198791503906 -6.842399597167969 -2.2807998657226562
Progress: 0 1.3005119562149048
   grad: 1.0 4.0 -0.7661924362182617 -0.7661924362182617 -0.7661924362182617
   grad: 2.0 9.0 -0.355377197265625 -0.1776885986328125 -0.08884429931640625
   grad: 3.0 16.0 19.31283187866211 6.437610626220703 2.1458702087402344
Progress: 1 1.1511896848678589
   grad: 1.0 4.0 -1.2657089233398438 -1.2657089233398438 -1.2657089233398438
   grad: 2.0 9.0 -6.879127502441406 -3.439563751220703 -1.7197818756103516
   grad: 3.0 16.0 0.4363288879394531 0.14544296264648438 0.048480987548828125
Progress: 2 0.0005876015638932586
   grad: 1.0 4.0 -0.9616022109985352 -0.9616022109985352 -0.9616022109985352
   grad: 2.0 9.0 -3.6181716918945312 -1.8090858459472656 -0.9045429229736328
   grad: 3.0 16.0 8.893535614013672 2.9645118713378906 0.9881706237792969
Progress: 3 0.2441202998161316
   grad: 1.0 4.0 -1.0341944694519043 -1.0341944694519043 -1.0341944694519043
   grad: 2.0 9.0 -4.918693542480469 -2.4593467712402344 -1.2296733856201172
   grad: 3.0 16.0 4.644985198974609 1.5483283996582031 0.5161094665527344
Progress: 4 0.0665922462940216
   grad: 1.0 4.0 -0.9341764450073242 -0.9341764450073242 -0.9341764450073242
   grad: 2.0 9.0 -4.105125427246094 -2.052562713623047 -1.0262813568115234
   grad: 3.0 16.0 6.320503234863281 2.1068344116210938 0.7022781372070312
Progress: 5 0.12329864501953125
   grad: 1.0 4.0 -0.9170389175415039 -0.9170389175415039 -0.9170389175415039
   grad: 2.0 9.0 -4.28326416015625 -2.141632080078125 -1.0708160400390625
   grad: 3.0 16.0 5.251155853271484 1.7503852844238281 0.5834617614746094
Progress: 6 0.08510690927505493
   grad: 1.0 4.0 -0.8638019561767578 -0.8638019561767578 -0.8638019561767578
   grad: 2.0 9.0 -4.0076751708984375 -2.0038375854492188 -1.0019187927246094
   grad: 3.0 16.0 5.4702301025390625 1.8234100341796875 0.6078033447265625
Progress: 7 0.09235622733831406
   grad: 1.0 4.0 -0.8297348022460938 -0.8297348022460938 -0.8297348022460938
   grad: 2.0 9.0 -3.9506378173828125 -1.9753189086914062 -0.9876594543457031
   grad: 3.0 16.0 5.100437164306641 1.7001457214355469 0.5667152404785156
Progress: 8 0.08029153943061829
   grad: 1.0 4.0 -0.7890243530273438 -0.7890243530273438 -0.7890243530273438
   grad: 2.0 9.0 -3.7990188598632812 -1.8995094299316406 -0.9497547149658203
   grad: 3.0 16.0 5.015052795410156 1.6716842651367188 0.5572280883789062
Progress: 9 0.07762578874826431
   grad: 1.0 4.0 -0.7535967826843262 -0.7535967826843262 -0.7535967826843262
   grad: 2.0 9.0 -3.6982879638671875 -1.8491439819335938 -0.9245719909667969
   grad: 3.0 16.0 4.807273864746094 1.6024246215820312 0.5341415405273438
Progress: 10 0.07132679969072342
   grad: 1.0 4.0 -0.7178177833557129 -0.7178177833557129 -0.7178177833557129
   grad: 2.0 9.0 -3.5804672241210938 -1.7902336120605469 -0.8951168060302734
   grad: 3.0 16.0 4.665927886962891 1.5553092956542969 0.5184364318847656
Progress: 11 0.06719408184289932
   grad: 1.0 4.0 -0.6842260360717773 -0.6842260360717773 -0.6842260360717773
   grad: 2.0 9.0 -3.4769287109375 -1.73846435546875 -0.869232177734375
   grad: 3.0 16.0 4.5028839111328125 1.5009613037109375 0.5003204345703125
Progress: 12 0.06258013099431992
   grad: 1.0 4.0 -0.6515631675720215 -0.6515631675720215 -0.6515631675720215
   grad: 2.0 9.0 -3.3728485107421875 -1.6864242553710938 -0.8432121276855469
   grad: 3.0 16.0 4.358791351318359 1.4529304504394531 0.4843101501464844
Progress: 13 0.058639079332351685
   grad: 1.0 4.0 -0.6203403472900391 -0.6203403472900391 -0.6203403472900391
   grad: 2.0 9.0 -3.2748870849609375 -1.6374435424804688 -0.8187217712402344
   grad: 3.0 16.0 4.214252471923828 1.4047508239746094 0.4682502746582031
Progress: 14 0.05481458082795143
   grad: 1.0 4.0 -0.5902442932128906 -0.5902442932128906 -0.5902442932128906
   grad: 2.0 9.0 -3.1796875 -1.58984375 -0.794921875
   grad: 3.0 16.0 4.078159332275391 1.3593864440917969 0.4531288146972656
Progress: 15 0.05133143067359924
   grad: 1.0 4.0 -0.5613541603088379 -0.5613541603088379 -0.5613541603088379
   grad: 2.0 9.0 -3.0886154174804688 -1.5443077087402344 -0.7721538543701172
   grad: 3.0 16.0 3.9459800720214844 1.3153266906738281 0.4384422302246094
Progress: 16 0.04805789887905121
   grad: 1.0 4.0 -0.5335664749145508 -0.5335664749145508 -0.5335664749145508
   grad: 2.0 9.0 -3.0008468627929688 -1.5004234313964844 -0.7502117156982422
   grad: 3.0 16.0 3.8194656372070312 1.2731552124023438 0.42438507080078125
Progress: 17 0.045025672763586044
   grad: 1.0 4.0 -0.5068626403808594 -0.5068626403808594 -0.5068626403808594
   grad: 2.0 9.0 -2.9165267944335938 -1.4582633972167969 -0.7291316986083984
   grad: 3.0 16.0 3.6975860595703125 1.2325286865234375 0.4108428955078125
Progress: 18 0.042197972536087036
   grad: 1.0 4.0 -0.48119115829467773 -0.48119115829467773 -0.48119115829467773
   grad: 2.0 9.0 -2.8354263305664062 -1.4177131652832031 -0.7088565826416016
   grad: 3.0 16.0 3.5805130004882812 1.1935043334960938 0.39783477783203125
Progress: 19 0.03956812620162964
   grad: 1.0 4.0 -0.45651721954345703 -0.45651721954345703 -0.45651721954345703
   grad: 2.0 9.0 -2.7574462890625 -1.37872314453125 -0.689361572265625
   grad: 3.0 16.0 3.4679031372070312 1.1559677124023438 0.38532257080078125
Progress: 20 0.037118371576070786
   grad: 1.0 4.0 -0.43280029296875 -0.43280029296875 -0.43280029296875
   grad: 2.0 9.0 -2.682464599609375 -1.3412322998046875 -0.6706161499023438
   grad: 3.0 16.0 3.3595848083496094 1.1198616027832031 0.3732872009277344
Progress: 21 0.03483583405613899
   grad: 1.0 4.0 -0.4100003242492676 -0.4100003242492676 -0.4100003242492676
   grad: 2.0 9.0 -2.6103363037109375 -1.3051681518554688 -0.6525840759277344
   grad: 3.0 16.0 3.2555580139160156 1.0851860046386719 0.3617286682128906
Progress: 22 0.03271190822124481
   grad: 1.0 4.0 -0.3880882263183594 -0.3880882263183594 -0.3880882263183594
   grad: 2.0 9.0 -2.541015625 -1.2705078125 -0.63525390625
   grad: 3.0 16.0 3.155445098876953 1.0518150329589844 0.3506050109863281
Progress: 23 0.030730968341231346
   grad: 1.0 4.0 -0.36702537536621094 -0.36702537536621094 -0.36702537536621094
   grad: 2.0 9.0 -2.474334716796875 -1.2371673583984375 -0.6185836791992188
   grad: 3.0 16.0 3.0591773986816406 1.0197257995605469 0.3399085998535156
Progress: 24 0.028884464874863625
   grad: 1.0 4.0 -0.34677839279174805 -0.34677839279174805 -0.34677839279174805
   grad: 2.0 9.0 -2.4102020263671875 -1.2051010131835938 -0.6025505065917969
   grad: 3.0 16.0 2.966686248779297 0.9888954162597656 0.3296318054199219
Progress: 25 0.027164282277226448
   grad: 1.0 4.0 -0.3273186683654785 -0.3273186683654785 -0.3273186683654785
   grad: 2.0 9.0 -2.3485641479492188 -1.1742820739746094 -0.5871410369873047
   grad: 3.0 16.0 2.8776626586914062 0.9592208862304688 0.31974029541015625
Progress: 26 0.025558464229106903
   grad: 1.0 4.0 -0.3086128234863281 -0.3086128234863281 -0.3086128234863281
   grad: 2.0 9.0 -2.2892532348632812 -1.1446266174316406 -0.5723133087158203
   grad: 3.0 16.0 2.79217529296875 0.93072509765625 0.31024169921875
Progress: 27 0.02406247705221176
   grad: 1.0 4.0 -0.2906355857849121 -0.2906355857849121 -0.2906355857849121
   grad: 2.0 9.0 -2.2322463989257812 -1.1161231994628906 -0.5580615997314453
   grad: 3.0 16.0 2.709880828857422 0.9032936096191406 0.3010978698730469
Progress: 28 0.022664980962872505
   grad: 1.0 4.0 -0.27335453033447266 -0.27335453033447266 -0.27335453033447266
   grad: 2.0 9.0 -2.177398681640625 -1.0886993408203125 -0.5443496704101562
   grad: 3.0 16.0 2.6308135986328125 0.8769378662109375 0.2923126220703125
Progress: 29 0.021361667662858963
   grad: 1.0 4.0 -0.2567453384399414 -0.2567453384399414 -0.2567453384399414
   grad: 2.0 9.0 -2.1246795654296875 -1.0623397827148438 -0.5311698913574219
   grad: 3.0 16.0 2.554767608642578 0.8515892028808594 0.2838630676269531
Progress: 30 0.02014455944299698
   grad: 1.0 4.0 -0.24078130722045898 -0.24078130722045898 -0.24078130722045898
   grad: 2.0 9.0 -2.0739593505859375 -1.0369796752929688 -0.5184898376464844
   grad: 3.0 16.0 2.481639862060547 0.8272132873535156 0.2757377624511719
Progress: 31 0.01900782808661461
   grad: 1.0 4.0 -0.22543811798095703 -0.22543811798095703 -0.22543811798095703
   grad: 2.0 9.0 -2.0251922607421875 -1.0125961303710938 -0.5062980651855469
   grad: 3.0 16.0 2.4112930297851562 0.8037643432617188 0.26792144775390625
Progress: 32 0.01794547587633133
   grad: 1.0 4.0 -0.21069002151489258 -0.21069002151489258 -0.21069002151489258
   grad: 2.0 9.0 -1.978271484375 -0.9891357421875 -0.49456787109375
   grad: 3.0 16.0 2.3437271118164062 0.7812423706054688 0.26041412353515625
Progress: 33 0.016953878104686737
   grad: 1.0 4.0 -0.1965165138244629 -0.1965165138244629 -0.1965165138244629
   grad: 2.0 9.0 -1.9331817626953125 -0.9665908813476562 -0.4832954406738281
   grad: 3.0 16.0 2.2787017822265625 0.7595672607421875 0.2531890869140625
Progress: 34 0.016026178374886513
   grad: 1.0 4.0 -0.1828932762145996 -0.1828932762145996 -0.1828932762145996
   grad: 2.0 9.0 -1.8897933959960938 -0.9448966979980469 -0.47244834899902344
   grad: 3.0 16.0 2.2161483764648438 0.7387161254882812 0.24623870849609375
Progress: 35 0.015158375725150108
   grad: 1.0 4.0 -0.1697988510131836 -0.1697988510131836 -0.1697988510131836
   grad: 2.0 9.0 -1.848052978515625 -0.9240264892578125 -0.46201324462890625
   grad: 3.0 16.0 2.1561012268066406 0.7187004089355469 0.23956680297851562
Progress: 36 0.014348063617944717
   grad: 1.0 4.0 -0.15721654891967773 -0.15721654891967773 -0.15721654891967773
   grad: 2.0 9.0 -1.8079376220703125 -0.9039688110351562 -0.4519844055175781
   grad: 3.0 16.0 2.098217010498047 0.6994056701660156 0.23313522338867188
Progress: 37 0.013588008470833302
   grad: 1.0 4.0 -0.14512109756469727 -0.14512109756469727 -0.14512109756469727
   grad: 2.0 9.0 -1.769317626953125 -0.8846588134765625 -0.44232940673828125
   grad: 3.0 16.0 2.0427017211914062 0.6809005737304688 0.22696685791015625
Progress: 38 0.01287848874926567
   grad: 1.0 4.0 -0.1334986686706543 -0.1334986686706543 -0.1334986686706543
   grad: 2.0 9.0 -1.732208251953125 -0.8661041259765625 -0.43305206298828125
   grad: 3.0 16.0 1.9892120361328125 0.6630706787109375 0.2210235595703125
Progress: 39 0.012212853878736496
   grad: 1.0 4.0 -0.12232685089111328 -0.12232685089111328 -0.12232685089111328
   grad: 2.0 9.0 -1.6964950561523438 -0.8482475280761719 -0.42412376403808594
   grad: 3.0 16.0 1.9378166198730469 0.6459388732910156 0.21531295776367188
Progress: 40 0.011589917354285717
   grad: 1.0 4.0 -0.11159133911132812 -0.11159133911132812 -0.11159133911132812
   grad: 2.0 9.0 -1.6621551513671875 -0.8310775756835938 -0.4155387878417969
   grad: 3.0 16.0 1.8884124755859375 0.6294708251953125 0.2098236083984375
Progress: 41 0.011006486602127552
   grad: 1.0 4.0 -0.10127449035644531 -0.10127449035644531 -0.10127449035644531
   grad: 2.0 9.0 -1.6291275024414062 -0.8145637512207031 -0.40728187561035156
   grad: 3.0 16.0 1.8408279418945312 0.6136093139648438 0.20453643798828125
Progress: 42 0.010458788834512234
   grad: 1.0 4.0 -0.09135770797729492 -0.09135770797729492 -0.09135770797729492
   grad: 2.0 9.0 -1.5973358154296875 -0.7986679077148438 -0.3993339538574219
   grad: 3.0 16.0 1.795166015625 0.598388671875 0.199462890625
Progress: 43 0.009946361184120178
   grad: 1.0 4.0 -0.08183002471923828 -0.08183002471923828 -0.08183002471923828
   grad: 2.0 9.0 -1.5667800903320312 -0.7833900451660156 -0.3916950225830078
   grad: 3.0 16.0 1.7511520385742188 0.5837173461914062 0.19457244873046875
Progress: 44 0.009464609436690807
   grad: 1.0 4.0 -0.07267189025878906 -0.07267189025878906 -0.07267189025878906
   grad: 2.0 9.0 -1.5373687744140625 -0.7686843872070312 -0.3843421936035156
   grad: 3.0 16.0 1.70892333984375 0.56964111328125 0.18988037109375
Progress: 45 0.009013638831675053
   grad: 1.0 4.0 -0.06387186050415039 -0.06387186050415039 -0.06387186050415039
   grad: 2.0 9.0 -1.50909423828125 -0.754547119140625 -0.3772735595703125
   grad: 3.0 16.0 1.6682395935058594 0.5560798645019531 0.18535995483398438
Progress: 46 0.008589577861130238
   grad: 1.0 4.0 -0.05541515350341797 -0.05541515350341797 -0.05541515350341797
   grad: 2.0 9.0 -1.48187255859375 -0.740936279296875 -0.3704681396484375
   grad: 3.0 16.0 1.6291351318359375 0.5430450439453125 0.1810150146484375
Progress: 47 0.00819160882383585
   grad: 1.0 4.0 -0.04728841781616211 -0.04728841781616211 -0.04728841781616211
   grad: 2.0 9.0 -1.4557037353515625 -0.7278518676757812 -0.3639259338378906
   grad: 3.0 16.0 1.5915412902832031 0.5305137634277344 0.17683792114257812
Progress: 48 0.007817912846803665
   grad: 1.0 4.0 -0.03947925567626953 -0.03947925567626953 -0.03947925567626953
   grad: 2.0 9.0 -1.4305191040039062 -0.7152595520019531 -0.35762977600097656
   grad: 3.0 16.0 1.555389404296875 0.518463134765625 0.172821044921875
Progress: 49 0.0074667781591415405
   grad: 1.0 4.0 -0.031975746154785156 -0.031975746154785156 -0.031975746154785156
   grad: 2.0 9.0 -1.4063034057617188 -0.7031517028808594 -0.3515758514404297
   grad: 3.0 16.0 1.5205764770507812 0.5068588256835938 0.16895294189453125
Progress: 50 0.007136274129152298
   grad: 1.0 4.0 -0.0247650146484375 -0.0247650146484375 -0.0247650146484375
   grad: 2.0 9.0 -1.3829879760742188 -0.6914939880371094 -0.3457469940185547
   grad: 3.0 16.0 1.4871711730957031 0.4957237243652344 0.16524124145507812
Progress: 51 0.006826166994869709
   grad: 1.0 4.0 -0.017837047576904297 -0.017837047576904297 -0.017837047576904297
   grad: 2.0 9.0 -1.3605804443359375 -0.6802902221679688 -0.3401451110839844
   grad: 3.0 16.0 1.4549674987792969 0.4849891662597656 0.16166305541992188
Progress: 52 0.0065337358973920345
   grad: 1.0 4.0 -0.011178970336914062 -0.011178970336914062 -0.011178970336914062
   grad: 2.0 9.0 -1.3389892578125 -0.66949462890625 -0.334747314453125
   grad: 3.0 16.0 1.4240341186523438 0.47467803955078125 0.15822601318359375
Progress: 53 0.006258867681026459
   grad: 1.0 4.0 -0.004782199859619141 -0.004782199859619141 -0.004782199859619141
   grad: 2.0 9.0 -1.3182373046875 -0.65911865234375 -0.329559326171875
   grad: 3.0 16.0 1.3942680358886719 0.4647560119628906 0.15491867065429688
Progress: 54 0.00599994882941246
   grad: 1.0 4.0 0.0013637542724609375 0.0013637542724609375 0.0013637542724609375
   grad: 2.0 9.0 -1.2982635498046875 -0.6491317749023438 -0.3245658874511719
   grad: 3.0 16.0 1.3656349182128906 0.4552116394042969 0.15173721313476562
Progress: 55 0.005756045691668987
   grad: 1.0 4.0 0.0072689056396484375 0.0072689056396484375 0.0072689056396484375
   grad: 2.0 9.0 -1.2790374755859375 -0.6395187377929688 -0.3197593688964844
   grad: 3.0 16.0 1.3381004333496094 0.4460334777832031 0.14867782592773438
Progress: 56 0.005526273977011442
   grad: 1.0 4.0 0.012943267822265625 0.012943267822265625 0.012943267822265625
   grad: 2.0 9.0 -1.2605514526367188 -0.6302757263183594 -0.3151378631591797
   grad: 3.0 16.0 1.3115959167480469 0.4371986389160156 0.14573287963867188
Progress: 57 0.005309517960995436
   grad: 1.0 4.0 0.01839447021484375 0.01839447021484375 0.01839447021484375
   grad: 2.0 9.0 -1.24273681640625 -0.621368408203125 -0.3106842041015625
   grad: 3.0 16.0 1.2861557006835938 0.42871856689453125 0.14290618896484375
Progress: 58 0.005105544812977314
   grad: 1.0 4.0 0.02363109588623047 0.02363109588623047 0.02363109588623047
   grad: 2.0 9.0 -1.2256240844726562 -0.6128120422363281 -0.30640602111816406
   grad: 3.0 16.0 1.2616424560546875 0.4205474853515625 0.1401824951171875
Progress: 59 0.004912782926112413
   grad: 1.0 4.0 0.028662681579589844 0.028662681579589844 0.028662681579589844
   grad: 2.0 9.0 -1.209136962890625 -0.6045684814453125 -0.30228424072265625
   grad: 3.0 16.0 1.2380905151367188 0.41269683837890625 0.13756561279296875
Progress: 60 0.004731074441224337
   grad: 1.0 4.0 0.03349590301513672 0.03349590301513672 0.03349590301513672
   grad: 2.0 9.0 -1.1932830810546875 -0.5966415405273438 -0.2983207702636719
   grad: 3.0 16.0 1.215362548828125 0.405120849609375 0.135040283203125
Progress: 61 0.004558969289064407
   grad: 1.0 4.0 0.038140296936035156 0.038140296936035156 0.038140296936035156
   grad: 2.0 9.0 -1.1779937744140625 -0.5889968872070312 -0.2944984436035156
   grad: 3.0 16.0 1.1935958862304688 0.39786529541015625 0.13262176513671875
Progress: 62 0.004397133365273476
   grad: 1.0 4.0 0.04260063171386719 0.04260063171386719 0.04260063171386719
   grad: 2.0 9.0 -1.1633148193359375 -0.5816574096679688 -0.2908287048339844
   grad: 3.0 16.0 1.1725502014160156 0.3908500671386719 0.13028335571289062
Progress: 63 0.004243438132107258
   grad: 1.0 4.0 0.046886444091796875 0.046886444091796875 0.046886444091796875
   grad: 2.0 9.0 -1.1491622924804688 -0.5745811462402344 -0.2872905731201172
   grad: 3.0 16.0 1.1524314880371094 0.3841438293457031 0.12804794311523438
Progress: 64 0.0040990691632032394
   grad: 1.0 4.0 0.05100250244140625 0.05100250244140625 0.05100250244140625
   grad: 2.0 9.0 -1.1355514526367188 -0.5677757263183594 -0.2838878631591797
   grad: 3.0 16.0 1.132965087890625 0.377655029296875 0.125885009765625
Progress: 65 0.003961758688092232
   grad: 1.0 4.0 0.05495643615722656 0.05495643615722656 0.05495643615722656
   grad: 2.0 9.0 -1.1224365234375 -0.56121826171875 -0.280609130859375
   grad: 3.0 16.0 1.114288330078125 0.371429443359375 0.123809814453125
Progress: 66 0.00383221753872931
   grad: 1.0 4.0 0.058753013610839844 0.058753013610839844 0.058753013610839844
   grad: 2.0 9.0 -1.1098251342773438 -0.5549125671386719 -0.27745628356933594
   grad: 3.0 16.0 1.0962982177734375 0.3654327392578125 0.1218109130859375
Progress: 67 0.003709474578499794
   grad: 1.0 4.0 0.06240081787109375 0.06240081787109375 0.06240081787109375
   grad: 2.0 9.0 -1.0976791381835938 -0.5488395690917969 -0.27441978454589844
   grad: 3.0 16.0 1.0790290832519531 0.3596763610839844 0.11989212036132812
Progress: 68 0.0035935300402343273
   grad: 1.0 4.0 0.0659027099609375 0.0659027099609375 0.0659027099609375
   grad: 2.0 9.0 -1.0859832763671875 -0.5429916381835938 -0.2714958190917969
   grad: 3.0 16.0 1.0623779296875 0.3541259765625 0.1180419921875
Progress: 69 0.0034834779798984528
   grad: 1.0 4.0 0.06926727294921875 0.06926727294921875 0.06926727294921875
   grad: 2.0 9.0 -1.0747299194335938 -0.5373649597167969 -0.26868247985839844
   grad: 3.0 16.0 1.0463447570800781 0.3487815856933594 0.11626052856445312
Progress: 70 0.003379127709195018
   grad: 1.0 4.0 0.07249927520751953 0.07249927520751953 0.07249927520751953
   grad: 2.0 9.0 -1.0638809204101562 -0.5319404602050781 -0.26597023010253906
   grad: 3.0 16.0 1.0308952331542969 0.3436317443847656 0.11454391479492188
Progress: 71 0.0032800771296024323
   grad: 1.0 4.0 0.07560253143310547 0.07560253143310547 0.07560253143310547
   grad: 2.0 9.0 -1.0534210205078125 -0.5267105102539062 -0.2633552551269531
   grad: 3.0 16.0 1.0161323547363281 0.3387107849121094 0.11290359497070312
Progress: 72 0.0031868054065853357
   grad: 1.0 4.0 0.0785818099975586 0.0785818099975586 0.0785818099975586
   grad: 2.0 9.0 -1.0433807373046875 -0.5216903686523438 -0.2608451843261719
   grad: 3.0 16.0 1.0018501281738281 0.3339500427246094 0.11131668090820312
Progress: 73 0.0030978508293628693
   grad: 1.0 4.0 0.08144283294677734 0.08144283294677734 0.08144283294677734
   grad: 2.0 9.0 -1.0336990356445312 -0.5168495178222656 -0.2584247589111328
   grad: 3.0 16.0 0.9881172180175781 0.3293724060058594 0.10979080200195312
Progress: 74 0.003013505134731531
   grad: 1.0 4.0 0.08419036865234375 0.08419036865234375 0.08419036865234375
   grad: 2.0 9.0 -1.0243682861328125 -0.5121841430664062 -0.2560920715332031
   grad: 3.0 16.0 0.9748992919921875 0.3249664306640625 0.1083221435546875
Progress: 75 0.0029334216378629208
   grad: 1.0 4.0 0.08682823181152344 0.08682823181152344 0.08682823181152344
   grad: 2.0 9.0 -1.0153884887695312 -0.5076942443847656 -0.2538471221923828
   grad: 3.0 16.0 0.9621620178222656 0.3207206726074219 0.10690689086914062
Progress: 76 0.002857270883396268
   grad: 1.0 4.0 0.08936119079589844 0.08936119079589844 0.08936119079589844
   grad: 2.0 9.0 -1.0067291259765625 -0.5033645629882812 -0.2516822814941406
   grad: 3.0 16.0 0.9499740600585938 0.31665802001953125 0.10555267333984375
Progress: 77 0.002785341814160347
   grad: 1.0 4.0 0.09179115295410156 0.09179115295410156 0.09179115295410156
   grad: 2.0 9.0 -0.9984054565429688 -0.4992027282714844 -0.2496013641357422
   grad: 3.0 16.0 0.9380950927734375 0.3126983642578125 0.1042327880859375
Progress: 78 0.0027161184698343277
   grad: 1.0 4.0 0.09412765502929688 0.09412765502929688 0.09412765502929688
   grad: 2.0 9.0 -0.9903488159179688 -0.4951744079589844 -0.2475872039794922
   grad: 3.0 16.0 0.9268341064453125 0.3089447021484375 0.1029815673828125
Progress: 79 0.002651300746947527
   grad: 1.0 4.0 0.09636688232421875 0.09636688232421875 0.09636688232421875
   grad: 2.0 9.0 -0.9826202392578125 -0.49131011962890625 -0.24565505981445312
   grad: 3.0 16.0 0.9158821105957031 0.3052940368652344 0.10176467895507812
Progress: 80 0.002589012496173382
   grad: 1.0 4.0 0.09851646423339844 0.09851646423339844 0.09851646423339844
   grad: 2.0 9.0 -0.97515869140625 -0.487579345703125 -0.2437896728515625
   grad: 3.0 16.0 0.9053764343261719 0.3017921447753906 0.10059738159179688
Progress: 81 0.002529958263039589
   grad: 1.0 4.0 0.10058116912841797 0.10058116912841797 0.10058116912841797
   grad: 2.0 9.0 -0.9679794311523438 -0.4839897155761719 -0.24199485778808594
   grad: 3.0 16.0 0.8952827453613281 0.2984275817871094 0.09947586059570312
Progress: 82 0.002473861677572131
   grad: 1.0 4.0 0.10256195068359375 0.10256195068359375 0.10256195068359375
   grad: 2.0 9.0 -0.9610595703125 -0.48052978515625 -0.240264892578125
   grad: 3.0 16.0 0.885498046875 0.295166015625 0.098388671875
Progress: 83 0.002420082688331604
   grad: 1.0 4.0 0.10446453094482422 0.10446453094482422 0.10446453094482422
   grad: 2.0 9.0 -0.9543685913085938 -0.4771842956542969 -0.23859214782714844
   grad: 3.0 16.0 0.87615966796875 0.29205322265625 0.09735107421875
Progress: 84 0.0023693079128861427
   grad: 1.0 4.0 0.10628890991210938 0.10628890991210938 0.10628890991210938
   grad: 2.0 9.0 -0.9479446411132812 -0.4739723205566406 -0.2369861602783203
   grad: 3.0 16.0 0.8671302795410156 0.2890434265136719 0.09634780883789062
Progress: 85 0.002320725005120039
   grad: 1.0 4.0 0.10803890228271484 0.10803890228271484 0.10803890228271484
   grad: 2.0 9.0 -0.941741943359375 -0.4708709716796875 -0.23543548583984375
   grad: 3.0 16.0 0.8584098815917969 0.2861366271972656 0.09537887573242188
Progress: 86 0.0022742825094610453
   grad: 1.0 4.0 0.10971832275390625 0.10971832275390625 0.10971832275390625
   grad: 2.0 9.0 -0.935760498046875 -0.4678802490234375 -0.23394012451171875
   grad: 3.0 16.0 0.8500328063964844 0.2833442687988281 0.09444808959960938
Progress: 87 0.0022301103454083204
   grad: 1.0 4.0 0.11133098602294922 0.11133098602294922 0.11133098602294922
   grad: 2.0 9.0 -0.92999267578125 -0.464996337890625 -0.2324981689453125
   grad: 3.0 16.0 0.8419990539550781 0.2806663513183594 0.09355545043945312
Progress: 88 0.002188155660405755
   grad: 1.0 4.0 0.11287593841552734 0.11287593841552734 0.11287593841552734
   grad: 2.0 9.0 -0.9244384765625 -0.46221923828125 -0.231109619140625
   grad: 3.0 16.0 0.8342056274414062 0.27806854248046875 0.09268951416015625
Progress: 89 0.002147836610674858
   grad: 1.0 4.0 0.11435985565185547 0.11435985565185547 0.11435985565185547
   grad: 2.0 9.0 -0.9190597534179688 -0.4595298767089844 -0.2297649383544922
   grad: 3.0 16.0 0.8267555236816406 0.2755851745605469 0.09186172485351562
Progress: 90 0.0021096442360430956
   grad: 1.0 4.0 0.11578083038330078 0.11578083038330078 0.11578083038330078
   grad: 2.0 9.0 -0.9138946533203125 -0.45694732666015625 -0.22847366333007812
   grad: 3.0 16.0 0.8195114135742188 0.27317047119140625 0.09105682373046875
Progress: 91 0.002072836272418499
   grad: 1.0 4.0 0.11714649200439453 0.11714649200439453 0.11714649200439453
   grad: 2.0 9.0 -0.9088973999023438 -0.4544486999511719 -0.22722434997558594
   grad: 3.0 16.0 0.8125762939453125 0.2708587646484375 0.0902862548828125
Progress: 92 0.002037901896983385
   grad: 1.0 4.0 0.1184549331665039 0.1184549331665039 0.1184549331665039
   grad: 2.0 9.0 -0.904083251953125 -0.4520416259765625 -0.22602081298828125
   grad: 3.0 16.0 0.8059158325195312 0.26863861083984375 0.08954620361328125
Progress: 93 0.0020046306308358908
   grad: 1.0 4.0 0.11970806121826172 0.11970806121826172 0.11970806121826172
   grad: 2.0 9.0 -0.899444580078125 -0.4497222900390625 -0.22486114501953125
   grad: 3.0 16.0 0.7994613647460938 0.26648712158203125 0.08882904052734375
Progress: 94 0.001972649712115526
   grad: 1.0 4.0 0.1209096908569336 0.1209096908569336 0.1209096908569336
   grad: 2.0 9.0 -0.89495849609375 -0.447479248046875 -0.2237396240234375
   grad: 3.0 16.0 0.7932472229003906 0.2644157409667969 0.08813858032226562
Progress: 95 0.0019421023316681385
   grad: 1.0 4.0 0.12206268310546875 0.12206268310546875 0.12206268310546875
   grad: 2.0 9.0 -0.890625 -0.4453125 -0.22265625
   grad: 3.0 16.0 0.7872734069824219 0.2624244689941406 0.08747482299804688
Progress: 96 0.0019129611318930984
   grad: 1.0 4.0 0.1231679916381836 0.1231679916381836 0.1231679916381836
   grad: 2.0 9.0 -0.8864517211914062 -0.4432258605957031 -0.22161293029785156
   grad: 3.0 16.0 0.7815399169921875 0.2605133056640625 0.0868377685546875
Progress: 97 0.0018851994536817074
   grad: 1.0 4.0 0.12422561645507812 0.12422561645507812 0.12422561645507812
   grad: 2.0 9.0 -0.8824310302734375 -0.44121551513671875 -0.22060775756835938
   grad: 3.0 16.0 0.7759437561035156 0.2586479187011719 0.08621597290039062
Progress: 98 0.0018582985503599048
   grad: 1.0 4.0 0.12524127960205078 0.12524127960205078 0.12524127960205078
   grad: 2.0 9.0 -0.8785247802734375 -0.43926239013671875 -0.21963119506835938
   grad: 3.0 16.0 0.7706222534179688 0.25687408447265625 0.08562469482421875
Progress: 99 0.001832897076383233
Predict (after training) 4 25.259321212768555

Process finished with exit code 0

小于不是小鱼呀

关注

2
点赞
踩
0

收藏

觉得还不错? 一键收藏
0
评论
PyTorch深度学习实践（刘二大人）---第四讲反向传播

D:\Conda_data\envs\pytorch\python.exe D:/PycharmProject/Pytorch学习笔记-刘二大人/4.反向传播-作业.py。在Pytorch进行,tensor（标量、向量、矩阵、高维数据），保存data、grad。矩阵《The Matrix Cookbook》有很多公式，比如求导等，可以直接查找。Nonlinear function: （后面讲）
复制链接

扫一扫