Skip to content

Commit 3ffe6aa

Browse files
author
xyliao
committed
finish fcn and fix some bugs
1 parent 38e2e34 commit 3ffe6aa

File tree

10 files changed

+401
-437
lines changed

10 files changed

+401
-437
lines changed

‎.gitignore‎

Lines changed: 1 addition & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -1,8 +1,2 @@
1-
__pycache__
2-
*.pth
31
.ipynb_checkpoints
4-
img
5-
chapter3_MLP/3_Neural_Network/.desktop
6-
.vscode
7-
data
8-
mnist
2+
.idea

‎README.md‎

Lines changed: 2 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -20,7 +20,7 @@ Learn Deep Learning with PyTorch
2020
- [Tensor和Variable](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter2_PyTorch-Basics/Tensor-and-Variable.ipynb)
2121
- [自动求导机制](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter2_PyTorch-Basics/autograd.ipynb)
2222
- [动态图与静态图](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter2_PyTorch-Basics/dynamic-graph.ipynb)
23-
23+
2424

2525
- Chapter 3: 神经网络
2626
- [线性模型与梯度下降](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter3_NN/linear-regression-gradient-descend.ipynb)
@@ -35,7 +35,6 @@ Learn Deep Learning with PyTorch
3535
- [RMSProp](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter3_NN/optimizer/rmsprop.ipynb)
3636
- [Adadelta](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter3_NN/optimizer/adadelta.ipynb)
3737
- [Adam](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter3_NN/optimizer/adam.ipynb)
38-
3938
- Chapter 4: 卷积神经网络
4039
- [PyTorch 中的卷积模块](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter4_CNN/basic_conv.ipynb)
4140
- [批标准化，batch normalization](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter4_CNN/batch-normalization.ipynb)
@@ -47,7 +46,6 @@ Learn Deep Learning with PyTorch
4746
- [数据增强](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter4_CNN/data-augumentation.ipynb)
4847
- [正则化](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter4_CNN/regularization.ipynb)
4948
- [学习率衰减](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter4_CNN/lr-decay.ipynb)
50-
5149
- Chapter 5: 循环神经网络
5250
- [循环神经网络模块：LSTM 和 GRU](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter5_RNN/pytorch-rnn.ipynb)
5351
- [使用 RNN 进行图像分类](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter5_RNN/rnn-for-image.ipynb)
@@ -56,24 +54,22 @@ Learn Deep Learning with PyTorch
5654
- [Word Embedding](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter5_RNN/nlp/word-embedding.ipynb)
5755
- [N-Gram 模型](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter5_RNN/nlp/n-gram.ipynb)
5856
- [Seq-LSTM 做词性预测](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter5_RNN/nlp/seq-lstm.ipynb)
59-
6057
- Chapter 6: 生成对抗网络
6158
- [自动编码器](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter6_GAN/autoencoder.ipynb)
6259
- [变分自动编码器](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter6_GAN/vae.ipynb)
6360
- [生成对抗网络](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter6_GAN/gan.ipynb)
6461
- 深度卷积对抗网络 (DCGANs) 生成人脸
65-
6662
- Chapter 7: 深度强化学习
6763
- [Q Learning](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter7_RL/q-learning-intro.ipynb)
6864
- [Open AI gym](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter7_RL/open_ai_gym.ipynb)
6965
- [Deep Q-networks](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter7_RL/dqn.ipynb)
70-
7166
- Chapter 8: PyTorch高级
7267
- [tensorboard 可视化](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter8_PyTorch-Advances/tensorboard.ipynb)
7368
- [灵活的数据读取介绍](https://github.com/SherlockLiao/code-of-learn-deep-learning-with-pytorch/blob/master/chapter8_PyTorch-Advances/data-io.ipynb)
7469
- autograd.function 的介绍
7570
- 数据并行和多 GPU
7671
- 使用 ONNX 转化为 Caffe2 模型
72+
- 如何部署训练好的神经网络
7773
- 打造属于自己的 PyTorch 的使用习惯
7874

7975
### part2: 深度学习的应用

‎chapter2_PyTorch-Basics/autograd.ipynb‎

Lines changed: 40 additions & 73 deletions
Original file line numberDiff line numberDiff line change
@@ -31,9 +31,7 @@
3131
{
3232
"cell_type": "code",
3333
"execution_count": 2,
34-
"metadata": {
35-
"collapsed": false
36-
},
34+
"metadata": {},
3735
"outputs": [
3836
{
3937
"name": "stdout",
@@ -74,9 +72,7 @@
7472
{
7573
"cell_type": "code",
7674
"execution_count": 3,
77-
"metadata": {
78-
"collapsed": false
79-
},
75+
"metadata": {},
8076
"outputs": [
8177
{
8278
"name": "stdout",
@@ -105,9 +101,7 @@
105101
{
106102
"cell_type": "code",
107103
"execution_count": 4,
108-
"metadata": {
109-
"collapsed": false
110-
},
104+
"metadata": {},
111105
"outputs": [],
112106
"source": [
113107
"x = Variable(torch.randn(10, 20), requires_grad=True)\n",
@@ -128,9 +122,7 @@
128122
{
129123
"cell_type": "code",
130124
"execution_count": 5,
131-
"metadata": {
132-
"collapsed": false
133-
},
125+
"metadata": {},
134126
"outputs": [
135127
{
136128
"name": "stdout",
@@ -174,9 +166,7 @@
174166
{
175167
"cell_type": "code",
176168
"execution_count": 6,
177-
"metadata": {
178-
"collapsed": false
179-
},
169+
"metadata": {},
180170
"outputs": [
181171
{
182172
"name": "stdout",
@@ -207,9 +197,7 @@
207197
{
208198
"cell_type": "code",
209199
"execution_count": 7,
210-
"metadata": {
211-
"collapsed": false
212-
},
200+
"metadata": {},
213201
"outputs": [
214202
{
215203
"name": "stdout",
@@ -271,9 +259,7 @@
271259
{
272260
"cell_type": "code",
273261
"execution_count": 8,
274-
"metadata": {
275-
"collapsed": false
276-
},
262+
"metadata": {},
277263
"outputs": [
278264
{
279265
"name": "stdout",
@@ -300,9 +286,7 @@
300286
{
301287
"cell_type": "code",
302288
"execution_count": 9,
303-
"metadata": {
304-
"collapsed": false
305-
},
289+
"metadata": {},
306290
"outputs": [
307291
{
308292
"name": "stdout",
@@ -361,9 +345,7 @@
361345
{
362346
"cell_type": "code",
363347
"execution_count": 10,
364-
"metadata": {
365-
"collapsed": false
366-
},
348+
"metadata": {},
367349
"outputs": [],
368350
"source": [
369351
"n.backward(torch.ones_like(n)) # 将 (w0, w1) 取成 (1, 1)"
@@ -372,9 +354,7 @@
372354
{
373355
"cell_type": "code",
374356
"execution_count": 11,
375-
"metadata": {
376-
"collapsed": false
377-
},
357+
"metadata": {},
378358
"outputs": [
379359
{
380360
"name": "stdout",
@@ -423,9 +403,7 @@
423403
{
424404
"cell_type": "code",
425405
"execution_count": 12,
426-
"metadata": {
427-
"collapsed": false
428-
},
406+
"metadata": {},
429407
"outputs": [
430408
{
431409
"name": "stdout",
@@ -447,9 +425,7 @@
447425
{
448426
"cell_type": "code",
449427
"execution_count": 13,
450-
"metadata": {
451-
"collapsed": false
452-
},
428+
"metadata": {},
453429
"outputs": [],
454430
"source": [
455431
"y.backward(retain_graph=True) # 设置 retain_graph 为 True 来保留计算图"
@@ -458,9 +434,7 @@
458434
{
459435
"cell_type": "code",
460436
"execution_count": 14,
461-
"metadata": {
462-
"collapsed": false
463-
},
437+
"metadata": {},
464438
"outputs": [
465439
{
466440
"name": "stdout",
@@ -491,9 +465,7 @@
491465
{
492466
"cell_type": "code",
493467
"execution_count": 16,
494-
"metadata": {
495-
"collapsed": false
496-
},
468+
"metadata": {},
497469
"outputs": [
498470
{
499471
"name": "stdout",
@@ -568,42 +540,41 @@
568540
"$$\n",
569541
"\\left[\n",
570542
"\\begin{matrix}\n",
571-
"4 & 6 \\\\\n",
572-
"3 & 9 \\\\\n",
543+
"4 & 3 \\\\\n",
544+
"2 & 6 \\\\\n",
573545
"\\end{matrix}\n",
574546
"\\right]\n",
575547
"$$"
576548
]
577549
},
578550
{
579551
"cell_type": "code",
580-
"execution_count": 17,
552+
"execution_count": 6,
581553
"metadata": {
582554
"collapsed": true
583555
},
584556
"outputs": [],
585557
"source": [
586-
"x = Variable(torch.FloatTensor([[2, 3]]), requires_grad=True)\n",
587-
"k = Variable(torch.zeros(1, 2))\n",
558+
"x = Variable(torch.FloatTensor([2, 3]), requires_grad=True)\n",
559+
"k = Variable(torch.zeros(2))\n",
588560
"\n",
589-
"k[0, 0] = x[0, 0] ** 2 + 3 * x[0 ,1]\n",
590-
"k[0, 1] = x[0, 1] ** 2 + 2 * x[0, 0]"
561+
"k[0] = x[0] ** 2 + 3 * x[1]\n",
562+
"k[1] = x[1] ** 2 + 2 * x[0]"
591563
]
592564
},
593565
{
594566
"cell_type": "code",
595-
"execution_count": 18,
596-
"metadata": {
597-
"collapsed": false
598-
},
567+
"execution_count": 7,
568+
"metadata": {},
599569
"outputs": [
600570
{
601571
"name": "stdout",
602572
"output_type": "stream",
603573
"text": [
604574
"Variable containing:\n",
605-
" 13 13\n",
606-
"[torch.FloatTensor of size 1x2]\n",
575+
" 13\n",
576+
" 13\n",
577+
"[torch.FloatTensor of size 2]\n",
607578
"\n"
608579
]
609580
}
@@ -614,37 +585,33 @@
614585
},
615586
{
616587
"cell_type": "code",
617-
"execution_count": 19,
618-
"metadata": {
619-
"collapsed": false
620-
},
588+
"execution_count": 8,
589+
"metadata": {},
621590
"outputs": [],
622591
"source": [
623592
"j = torch.zeros(2, 2)\n",
624593
"\n",
625-
"k.backward(torch.FloatTensor([[1, 0]]), retain_graph=True)\n",
626-
"j[:, 0] = x.grad.data\n",
594+
"k.backward(torch.FloatTensor([1, 0]), retain_graph=True)\n",
595+
"j[0] = x.grad.data\n",
627596
"\n",
628-
"m.grad.data.zero_() # 归零之前求得的梯度\n",
597+
"x.grad.data.zero_() # 归零之前求得的梯度\n",
629598
"\n",
630-
"k.backward(torch.FloatTensor([[0, 1]]))\n",
631-
"j[:, 1] = x.grad.data"
599+
"k.backward(torch.FloatTensor([0, 1]))\n",
600+
"j[1] = x.grad.data"
632601
]
633602
},
634603
{
635604
"cell_type": "code",
636-
"execution_count": 20,
637-
"metadata": {
638-
"collapsed": false
639-
},
605+
"execution_count": 9,
606+
"metadata": {},
640607
"outputs": [
641608
{
642609
"name": "stdout",
643610
"output_type": "stream",
644611
"text": [
645612
"\n",
646-
" 4 6\n",
647-
" 3 9\n",
613+
" 4 3\n",
614+
" 2 6\n",
648615
"[torch.FloatTensor of size 2x2]\n",
649616
"\n"
650617
]
@@ -664,9 +631,9 @@
664631
],
665632
"metadata": {
666633
"kernelspec": {
667-
"display_name": "mx",
634+
"display_name": "Python 3",
668635
"language": "python",
669-
"name": "mx"
636+
"name": "python3"
670637
},
671638
"language_info": {
672639
"codemirror_mode": {
@@ -678,7 +645,7 @@
678645
"name": "python",
679646
"nbconvert_exporter": "python",
680647
"pygments_lexer": "ipython3",
681-
"version": "3.6.0"
648+
"version": "3.6.3"
682649
}
683650
},
684651
"nbformat": 4,

‎chapter3_NN/nn-sequential-module.ipynb‎

Lines changed: 57 additions & 105 deletions
Large diffs are not rendered by default.

0 commit comments

Comments
 (0)