andre/PyTorch-Tutorial
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add notebooks

Browse Source
This commit is contained in:
Fuyang Liu
2017-06-22 23:05:34 +02:00
parent 1adcbb999f
commit cdd6704b86
19 changed files with 5775 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

397
tutorial-contents-notebooks/201_torch_numpy.ipynb Normal file
View File

@ -0,0 +1,397 @@
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# 201 Torch and Numpy\n",
"\n",
"View more, visit my tutorial page: https://morvanzhou.github.io/tutorials/\n",
"My Youtube Channel: https://www.youtube.com/user/MorvanZhou\n",
"\n",
"Dependencies:\n",
"* torch: 0.1.11\n",
"* numpy\n",
"\n",
"Details about math operation in torch can be found in: http://pytorch.org/docs/torch.html#math-operations\n"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"import torch\n",
"import numpy as np"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"numpy array: [[0 1 2]\n",
" [3 4 5]] \n",
"torch tensor: \n",
" 0 1 2\n",
" 3 4 5\n",
"[torch.LongTensor of size 2x3]\n",
" \n",
"tensor to array: [[0 1 2]\n",
" [3 4 5]]\n"
]
}
],
"source": [
"# convert numpy to tensor or vise versa\n",
"np_data = np.arange(6).reshape((2, 3))\n",
"torch_data = torch.from_numpy(np_data)\n",
"tensor2array = torch_data.numpy()\n",
"print(\n",
" '\\nnumpy array:', np_data, # [[0 1 2], [3 4 5]]\n",
" '\\ntorch tensor:', torch_data, # 0 1 2 \\n 3 4 5 [torch.LongTensor of size 2x3]\n",
" '\\ntensor to array:', tensor2array, # [[0 1 2], [3 4 5]]\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"abs \n",
"numpy: [1 2 1 2] \n",
"torch: \n",
" 1\n",
" 2\n",
" 1\n",
" 2\n",
"[torch.FloatTensor of size 4]\n",
"\n"
]
}
],
"source": [
"# abs\n",
"data = [-1, -2, 1, 2]\n",
"tensor = torch.FloatTensor(data) # 32-bit floating point\n",
"print(\n",
" '\\nabs',\n",
" '\\nnumpy: ', np.abs(data), # [1 2 1 2]\n",
" '\\ntorch: ', torch.abs(tensor) # [1 2 1 2]\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"\n",
" 1\n",
" 2\n",
" 1\n",
" 2\n",
"[torch.FloatTensor of size 4]"
]
},
"execution_count": 4,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"tensor.abs()"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"sin \n",
"numpy: [-0.84147098 -0.90929743 0.84147098 0.90929743] \n",
"torch: \n",
"-0.8415\n",
"-0.9093\n",
" 0.8415\n",
" 0.9093\n",
"[torch.FloatTensor of size 4]\n",
"\n"
]
}
],
"source": [
"# sin\n",
"print(\n",
" '\\nsin',\n",
" '\\nnumpy: ', np.sin(data), # [-0.84147098 -0.90929743 0.84147098 0.90929743]\n",
" '\\ntorch: ', torch.sin(tensor) # [-0.8415 -0.9093 0.8415 0.9093]\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"\n",
" 0.2689\n",
" 0.1192\n",
" 0.7311\n",
" 0.8808\n",
"[torch.FloatTensor of size 4]"
]
},
"execution_count": 6,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"tensor.sigmoid()"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"\n",
" 0.3679\n",
" 0.1353\n",
" 2.7183\n",
" 7.3891\n",
"[torch.FloatTensor of size 4]"
]
},
"execution_count": 7,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"tensor.exp()"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"mean \n",
"numpy: 0.0 \n",
"torch: 0.0\n"
]
}
],
"source": [
"# mean\n",
"print(\n",
" '\\nmean',\n",
" '\\nnumpy: ', np.mean(data), # 0.0\n",
" '\\ntorch: ', torch.mean(tensor) # 0.0\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"matrix multiplication (matmul) \n",
"numpy: [[ 7 10]\n",
" [15 22]] \n",
"torch: \n",
" 7 10\n",
" 15 22\n",
"[torch.FloatTensor of size 2x2]\n",
"\n"
]
}
],
"source": [
"# matrix multiplication\n",
"data = [[1,2], [3,4]]\n",
"tensor = torch.FloatTensor(data) # 32-bit floating point\n",
"# correct method\n",
"print(\n",
" '\\nmatrix multiplication (matmul)',\n",
" '\\nnumpy: ', np.matmul(data, data), # [[7, 10], [15, 22]]\n",
" '\\ntorch: ', torch.mm(tensor, tensor) # [[7, 10], [15, 22]]\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 14,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"matrix multiplication (dot) \n",
"numpy: [[ 7 10]\n",
" [15 22]] \n",
"torch: 30.0\n"
]
}
],
"source": [
"# incorrect method\n",
"data = np.array(data)\n",
"print(\n",
" '\\nmatrix multiplication (dot)',\n",
" '\\nnumpy: ', data.dot(data), # [[7, 10], [15, 22]]\n",
" '\\ntorch: ', tensor.dot(tensor) # this will convert tensor to [1,2,3,4], you'll get 30.0\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Note that:\n",
"\n",
"torch.dot(tensor1, tensor2) → float\n",
"\n",
"Computes the dot product (inner product) of two tensors. Both tensors are treated as 1-D vectors."
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"\n",
" 7 10\n",
" 15 22\n",
"[torch.FloatTensor of size 2x2]"
]
},
"execution_count": 11,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"tensor.mm(tensor)"
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"\n",
" 1 4\n",
" 9 16\n",
"[torch.FloatTensor of size 2x2]"
]
},
"execution_count": 12,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"tensor * tensor"
]
},
{
"cell_type": "code",
"execution_count": 13,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"30.0"
]
},
"execution_count": 13,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"tensor.dot(tensor)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.5.2"
}
},
"nbformat": 4,
"nbformat_minor": 2
}

293
tutorial-contents-notebooks/202_variable.ipynb Normal file
View File

@ -0,0 +1,293 @@
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# 202 Variable\n",
"\n",
"View more, visit my tutorial page: https://morvanzhou.github.io/tutorials/\n",
"My Youtube Channel: https://www.youtube.com/user/MorvanZhou\n",
"\n",
"Dependencies:\n",
"* torch: 0.1.11\n",
"\n",
"Variable in torch is to build a computational graph,\n",
"but this graph is dynamic compared with a static graph in Tensorflow or Theano.\n",
"So torch does not have placeholder, torch can just pass variable to the computational graph.\n"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"import torch\n",
"from torch.autograd import Variable"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
" 1 2\n",
" 3 4\n",
"[torch.FloatTensor of size 2x2]\n",
"\n",
"Variable containing:\n",
" 1 2\n",
" 3 4\n",
"[torch.FloatTensor of size 2x2]\n",
"\n"
]
}
],
"source": [
"tensor = torch.FloatTensor([[1,2],[3,4]]) # build a tensor\n",
"variable = Variable(tensor, requires_grad=True) # build a variable, usually for compute gradients\n",
"\n",
"print(tensor) # [torch.FloatTensor of size 2x2]\n",
"print(variable) # [torch.FloatTensor of size 2x2]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Till now the tensor and variable seem the same.\n",
"\n",
"However, the variable is a part of the graph, it's a part of the auto-gradient.\n"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"7.5\n",
"Variable containing:\n",
" 7.5000\n",
"[torch.FloatTensor of size 1]\n",
"\n"
]
}
],
"source": [
"t_out = torch.mean(tensor*tensor) # x^2\n",
"v_out = torch.mean(variable*variable) # x^2\n",
"print(t_out)\n",
"print(v_out)"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"v_out.backward() # backpropagation from v_out"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"$$ v_{out} = {{1} \\over {4}} sum(variable^2) $$\n",
"\n",
"the gradients w.r.t the variable, \n",
"\n",
"$$ {d(v_{out}) \\over d(variable)} = {{1} \\over {4}} 2 variable = {variable \\over 2}$$\n",
"\n",
"let's check the result pytorch calculated for us below:"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"Variable containing:\n",
" 0.5000 1.0000\n",
" 1.5000 2.0000\n",
"[torch.FloatTensor of size 2x2]"
]
},
"execution_count": 5,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"variable.grad"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"Variable containing:\n",
" 1 2\n",
" 3 4\n",
"[torch.FloatTensor of size 2x2]"
]
},
"execution_count": 6,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"variable # this is data in variable format"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"\n",
" 1 2\n",
" 3 4\n",
"[torch.FloatTensor of size 2x2]"
]
},
"execution_count": 7,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"variable.data # this is data in tensor format"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([[ 1., 2.],\n",
" [ 3., 4.]], dtype=float32)"
]
},
"execution_count": 8,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"variable.data.numpy() # numpy format"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Note that we did `.backward()` on `v_out` but `variable` has been assigned new values on it's `grad`.\n",
"\n",
"As this line \n",
"```\n",
"v_out = torch.mean(variable*variable)\n",
"``` \n",
"will make a new variable `v_out` and connect it with `variable` in computation graph."
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"torch.autograd.variable.Variable"
]
},
"execution_count": 9,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"type(v_out)"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"torch.FloatTensor"
]
},
"execution_count": 10,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"type(v_out.data)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.5.2"
}
},
"nbformat": 4,
"nbformat_minor": 2
}

164
tutorial-contents-notebooks/203_activation.ipynb Normal file
View File

File diff suppressed because one or more lines are too long

302
tutorial-contents-notebooks/301_regression.ipynb Normal file
View File

File diff suppressed because one or more lines are too long

315
tutorial-contents-notebooks/302_classification.ipynb Normal file
View File

File diff suppressed because one or more lines are too long

133
tutorial-contents-notebooks/303_build_nn_quickly.ipynb Normal file
View File

@ -0,0 +1,133 @@
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# 303 Build NN Quickly\n",
"\n",
"View more, visit my tutorial page: https://morvanzhou.github.io/tutorials/\n",
"My Youtube Channel: https://www.youtube.com/user/MorvanZhou\n",
"\n",
"Dependencies:\n",
"* torch: 0.1.11"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"import torch\n",
"import torch.nn.functional as F"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"# replace following class code with an easy sequential network\n",
"class Net(torch.nn.Module):\n",
" def __init__(self, n_feature, n_hidden, n_output):\n",
" super(Net, self).__init__()\n",
" self.hidden = torch.nn.Linear(n_feature, n_hidden) # hidden layer\n",
" self.predict = torch.nn.Linear(n_hidden, n_output) # output layer\n",
"\n",
" def forward(self, x):\n",
" x = F.relu(self.hidden(x)) # activation function for hidden layer\n",
" x = self.predict(x) # linear output\n",
" return x"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"net1 = Net(1, 10, 1)"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"# easy and fast way to build your network\n",
"net2 = torch.nn.Sequential(\n",
" torch.nn.Linear(1, 10),\n",
" torch.nn.ReLU(),\n",
" torch.nn.Linear(10, 1)\n",
")\n"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Net (\n",
" (hidden): Linear (1 -> 10)\n",
" (predict): Linear (10 -> 1)\n",
")\n",
"Sequential (\n",
" (0): Linear (1 -> 10)\n",
" (1): ReLU ()\n",
" (2): Linear (10 -> 1)\n",
")\n"
]
}
],
"source": [
"print(net1) # net1 architecture\n",
"print(net2) # net2 architecture"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.5.2"
}
},
"nbformat": 4,
"nbformat_minor": 2
}

204
tutorial-contents-notebooks/304_save_reload.ipynb Normal file
View File

File diff suppressed because one or more lines are too long

177
tutorial-contents-notebooks/305_batch_train.ipynb Normal file
View File

@ -0,0 +1,177 @@
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# 305 Batch Train\n",
"\n",
"View more, visit my tutorial page: https://morvanzhou.github.io/tutorials/\n",
"My Youtube Channel: https://www.youtube.com/user/MorvanZhou\n",
"\n",
"Dependencies:\n",
"* torch: 0.1.11"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"<torch._C.Generator at 0x7faffc159918>"
]
},
"execution_count": 1,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"import torch\n",
"import torch.utils.data as Data\n",
"\n",
"torch.manual_seed(1) # reproducible"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"BATCH_SIZE = 5\n",
"# BATCH_SIZE = 8"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"x = torch.linspace(1, 10, 10) # this is x data (torch tensor)\n",
"y = torch.linspace(10, 1, 10) # this is y data (torch tensor)\n"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"torch_dataset = Data.TensorDataset(data_tensor=x, target_tensor=y)\n",
"loader = Data.DataLoader(\n",
" dataset=torch_dataset, # torch TensorDataset format\n",
" batch_size=BATCH_SIZE, # mini batch size\n",
" shuffle=True, # random shuffle for training\n",
" num_workers=2, # subprocesses for loading data\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Epoch: 0 | Step: 0 | batch x: [ 6. 7. 2. 3. 1.] | batch y: [ 5. 4. 9. 8. 10.]\n",
"Epoch: 0 | Step: 1 | batch x: [ 9. 10. 4. 8. 5.] | batch y: [ 2. 1. 7. 3. 6.]\n",
"Epoch: 1 | Step: 0 | batch x: [ 3. 4. 2. 9. 10.] | batch y: [ 8. 7. 9. 2. 1.]\n",
"Epoch: 1 | Step: 1 | batch x: [ 1. 7. 8. 5. 6.] | batch y: [ 10. 4. 3. 6. 5.]\n",
"Epoch: 2 | Step: 0 | batch x: [ 3. 9. 2. 6. 7.] | batch y: [ 8. 2. 9. 5. 4.]\n",
"Epoch: 2 | Step: 1 | batch x: [ 10. 4. 8. 1. 5.] | batch y: [ 1. 7. 3. 10. 6.]\n"
]
}
],
"source": [
"for epoch in range(3): # train entire dataset 3 times\n",
" for step, (batch_x, batch_y) in enumerate(loader): # for each training step\n",
" # train your data...\n",
" print('Epoch: ', epoch, '| Step: ', step, '| batch x: ',\n",
" batch_x.numpy(), '| batch y: ', batch_y.numpy())\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Suppose a different batch size that cannot be fully divided by the number of data entreis:"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Epoch: 0 | Step: 0 | batch x: [ 3. 10. 9. 4. 7. 8. 2. 1.] | batch y: [ 8. 1. 2. 7. 4. 3. 9. 10.]\n",
"Epoch: 0 | Step: 1 | batch x: [ 5. 6.] | batch y: [ 6. 5.]\n",
"Epoch: 1 | Step: 0 | batch x: [ 4. 8. 3. 2. 1. 10. 5. 6.] | batch y: [ 7. 3. 8. 9. 10. 1. 6. 5.]\n",
"Epoch: 1 | Step: 1 | batch x: [ 7. 9.] | batch y: [ 4. 2.]\n",
"Epoch: 2 | Step: 0 | batch x: [ 6. 2. 4. 10. 9. 3. 8. 5.] | batch y: [ 5. 9. 7. 1. 2. 8. 3. 6.]\n",
"Epoch: 2 | Step: 1 | batch x: [ 7. 1.] | batch y: [ 4. 10.]\n"
]
}
],
"source": [
"BATCH_SIZE = 8\n",
"loader = Data.DataLoader(\n",
" dataset=torch_dataset, # torch TensorDataset format\n",
" batch_size=BATCH_SIZE, # mini batch size\n",
" shuffle=True, # random shuffle for training\n",
" num_workers=2, # subprocesses for loading data\n",
")\n",
"for epoch in range(3): # train entire dataset 3 times\n",
" for step, (batch_x, batch_y) in enumerate(loader): # for each training step\n",
" # train your data...\n",
" print('Epoch: ', epoch, '| Step: ', step, '| batch x: ',\n",
" batch_x.numpy(), '| batch y: ', batch_y.numpy())"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.5.2"
}
},
"nbformat": 4,
"nbformat_minor": 2
}

295
tutorial-contents-notebooks/306_optimizer.ipynb Normal file
View File

File diff suppressed because one or more lines are too long

529
tutorial-contents-notebooks/401_CNN.ipynb Normal file
View File

File diff suppressed because one or more lines are too long

318
tutorial-contents-notebooks/402_RNN.ipynb Normal file
View File

File diff suppressed because one or more lines are too long

264
tutorial-contents-notebooks/403_RNN_regressor.ipynb Normal file
View File

File diff suppressed because one or more lines are too long

407
tutorial-contents-notebooks/404_autoencoder.ipynb Normal file
View File

File diff suppressed because one or more lines are too long

433
tutorial-contents-notebooks/405_DQN_Reinforcement_learning.ipynb Normal file
View File

@ -0,0 +1,433 @@
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# 405 DQN Reinforcement Learning\n",
"\n",
"View more, visit my tutorial page: https://morvanzhou.github.io/tutorials/\n",
"My Youtube Channel: https://www.youtube.com/user/MorvanZhou\n",
"More about Reinforcement learning: https://morvanzhou.github.io/tutorials/machine-learning/reinforcement-learning/\n",
"\n",
"Dependencies:\n",
"* torch: 0.1.11\n",
"* gym: 0.8.1\n",
"* numpy"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"import torch\n",
"import torch.nn as nn\n",
"from torch.autograd import Variable\n",
"import torch.nn.functional as F\n",
"import numpy as np\n",
"import gym"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"[2017-06-20 22:23:40,418] Making new env: CartPole-v0\n"
]
}
],
"source": [
"# Hyper Parameters\n",
"BATCH_SIZE = 32\n",
"LR = 0.01 # learning rate\n",
"EPSILON = 0.9 # greedy policy\n",
"GAMMA = 0.9 # reward discount\n",
"TARGET_REPLACE_ITER = 100 # target update frequency\n",
"MEMORY_CAPACITY = 2000\n",
"env = gym.make('CartPole-v0')\n",
"env = env.unwrapped\n",
"N_ACTIONS = env.action_space.n\n",
"N_STATES = env.observation_space.shape[0]"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"class Net(nn.Module):\n",
" def __init__(self, ):\n",
" super(Net, self).__init__()\n",
" self.fc1 = nn.Linear(N_STATES, 10)\n",
" self.fc1.weight.data.normal_(0, 0.1) # initialization\n",
" self.out = nn.Linear(10, N_ACTIONS)\n",
" self.out.weight.data.normal_(0, 0.1) # initialization\n",
"\n",
" def forward(self, x):\n",
" x = self.fc1(x)\n",
" x = F.relu(x)\n",
" actions_value = self.out(x)\n",
" return actions_value"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"class DQN(object):\n",
" def __init__(self):\n",
" self.eval_net, self.target_net = Net(), Net()\n",
"\n",
" self.learn_step_counter = 0 # for target updating\n",
" self.memory_counter = 0 # for storing memory\n",
" self.memory = np.zeros((MEMORY_CAPACITY, N_STATES * 2 + 2)) # initialize memory\n",
" self.optimizer = torch.optim.Adam(self.eval_net.parameters(), lr=LR)\n",
" self.loss_func = nn.MSELoss()\n",
"\n",
" def choose_action(self, x):\n",
" x = Variable(torch.unsqueeze(torch.FloatTensor(x), 0))\n",
" # input only one sample\n",
" if np.random.uniform() < EPSILON: # greedy\n",
" actions_value = self.eval_net.forward(x)\n",
" action = torch.max(actions_value, 1)[1].data.numpy()[0, 0] # return the argmax\n",
" else: # random\n",
" action = np.random.randint(0, N_ACTIONS)\n",
" return action\n",
"\n",
" def store_transition(self, s, a, r, s_):\n",
" transition = np.hstack((s, [a, r], s_))\n",
" # replace the old memory with new memory\n",
" index = self.memory_counter % MEMORY_CAPACITY\n",
" self.memory[index, :] = transition\n",
" self.memory_counter += 1\n",
"\n",
" def learn(self):\n",
" # target parameter update\n",
" if self.learn_step_counter % TARGET_REPLACE_ITER == 0:\n",
" self.target_net.load_state_dict(self.eval_net.state_dict())\n",
" self.learn_step_counter += 1\n",
"\n",
" # sample batch transitions\n",
" sample_index = np.random.choice(MEMORY_CAPACITY, BATCH_SIZE)\n",
" b_memory = self.memory[sample_index, :]\n",
" b_s = Variable(torch.FloatTensor(b_memory[:, :N_STATES]))\n",
" b_a = Variable(torch.LongTensor(b_memory[:, N_STATES:N_STATES+1].astype(int)))\n",
" b_r = Variable(torch.FloatTensor(b_memory[:, N_STATES+1:N_STATES+2]))\n",
" b_s_ = Variable(torch.FloatTensor(b_memory[:, -N_STATES:]))\n",
"\n",
" # q_eval w.r.t the action in experience\n",
" q_eval = self.eval_net(b_s).gather(1, b_a) # shape (batch, 1)\n",
" q_next = self.target_net(b_s_).detach() # detach from graph, don't backpropagate\n",
" q_target = b_r + GAMMA * q_next.max(1)[0] # shape (batch, 1)\n",
" loss = self.loss_func(q_eval, q_target)\n",
"\n",
" self.optimizer.zero_grad()\n",
" loss.backward()\n",
" self.optimizer.step()"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"dqn = DQN()"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"\n",
"Collecting experience...\n",
"Ep: 201 | Ep_r: 1.59\n",
"Ep: 202 | Ep_r: 4.18\n",
"Ep: 203 | Ep_r: 2.73\n",
"Ep: 204 | Ep_r: 1.97\n",
"Ep: 205 | Ep_r: 1.18\n",
"Ep: 206 | Ep_r: 0.86\n",
"Ep: 207 | Ep_r: 2.88\n",
"Ep: 208 | Ep_r: 1.63\n",
"Ep: 209 | Ep_r: 3.91\n",
"Ep: 210 | Ep_r: 3.6\n",
"Ep: 211 | Ep_r: 0.98\n",
"Ep: 212 | Ep_r: 3.85\n",
"Ep: 213 | Ep_r: 1.81\n",
"Ep: 214 | Ep_r: 2.32\n",
"Ep: 215 | Ep_r: 3.75\n",
"Ep: 216 | Ep_r: 3.53\n",
"Ep: 217 | Ep_r: 4.75\n",
"Ep: 218 | Ep_r: 2.4\n",
"Ep: 219 | Ep_r: 0.64\n",
"Ep: 220 | Ep_r: 1.15\n",
"Ep: 221 | Ep_r: 2.3\n",
"Ep: 222 | Ep_r: 7.37\n",
"Ep: 223 | Ep_r: 1.25\n",
"Ep: 224 | Ep_r: 5.02\n",
"Ep: 225 | Ep_r: 10.29\n",
"Ep: 226 | Ep_r: 17.54\n",
"Ep: 227 | Ep_r: 36.2\n",
"Ep: 228 | Ep_r: 6.61\n",
"Ep: 229 | Ep_r: 10.04\n",
"Ep: 230 | Ep_r: 55.19\n",
"Ep: 231 | Ep_r: 10.03\n",
"Ep: 232 | Ep_r: 13.25\n",
"Ep: 233 | Ep_r: 8.75\n",
"Ep: 234 | Ep_r: 3.83\n",
"Ep: 235 | Ep_r: -0.92\n",
"Ep: 236 | Ep_r: 5.12\n",
"Ep: 237 | Ep_r: 3.56\n",
"Ep: 238 | Ep_r: 5.69\n",
"Ep: 239 | Ep_r: 8.43\n",
"Ep: 240 | Ep_r: 29.27\n",
"Ep: 241 | Ep_r: 17.95\n",
"Ep: 242 | Ep_r: 44.77\n",
"Ep: 243 | Ep_r: 98.0\n",
"Ep: 244 | Ep_r: 38.78\n",
"Ep: 245 | Ep_r: 45.02\n",
"Ep: 246 | Ep_r: 27.73\n",
"Ep: 247 | Ep_r: 36.96\n",
"Ep: 248 | Ep_r: 48.98\n",
"Ep: 249 | Ep_r: 111.36\n",
"Ep: 250 | Ep_r: 95.61\n",
"Ep: 251 | Ep_r: 149.77\n",
"Ep: 252 | Ep_r: 29.96\n",
"Ep: 253 | Ep_r: 2.79\n",
"Ep: 254 | Ep_r: 20.1\n",
"Ep: 255 | Ep_r: 24.25\n",
"Ep: 256 | Ep_r: 3074.75\n",
"Ep: 257 | Ep_r: 1258.49\n",
"Ep: 258 | Ep_r: 127.39\n",
"Ep: 259 | Ep_r: 283.46\n",
"Ep: 260 | Ep_r: 166.96\n",
"Ep: 261 | Ep_r: 101.71\n",
"Ep: 262 | Ep_r: 63.45\n",
"Ep: 263 | Ep_r: 288.94\n",
"Ep: 264 | Ep_r: 130.49\n",
"Ep: 265 | Ep_r: 207.05\n",
"Ep: 266 | Ep_r: 183.71\n",
"Ep: 267 | Ep_r: 142.75\n",
"Ep: 268 | Ep_r: 126.53\n",
"Ep: 269 | Ep_r: 310.79\n",
"Ep: 270 | Ep_r: 863.2\n",
"Ep: 271 | Ep_r: 365.12\n",
"Ep: 272 | Ep_r: 659.52\n",
"Ep: 273 | Ep_r: 103.98\n",
"Ep: 274 | Ep_r: 554.83\n",
"Ep: 275 | Ep_r: 246.01\n",
"Ep: 276 | Ep_r: 332.23\n",
"Ep: 277 | Ep_r: 323.35\n",
"Ep: 278 | Ep_r: 278.71\n",
"Ep: 279 | Ep_r: 613.6\n",
"Ep: 280 | Ep_r: 152.21\n",
"Ep: 281 | Ep_r: 402.02\n",
"Ep: 282 | Ep_r: 351.4\n",
"Ep: 283 | Ep_r: 115.87\n",
"Ep: 284 | Ep_r: 163.26\n",
"Ep: 285 | Ep_r: 631.0\n",
"Ep: 286 | Ep_r: 263.47\n",
"Ep: 287 | Ep_r: 511.21\n",
"Ep: 288 | Ep_r: 337.18\n",
"Ep: 289 | Ep_r: 819.76\n",
"Ep: 290 | Ep_r: 190.83\n",
"Ep: 291 | Ep_r: 442.98\n",
"Ep: 292 | Ep_r: 537.24\n",
"Ep: 293 | Ep_r: 1101.12\n",
"Ep: 294 | Ep_r: 178.42\n",
"Ep: 295 | Ep_r: 225.61\n",
"Ep: 296 | Ep_r: 252.62\n",
"Ep: 297 | Ep_r: 617.5\n",
"Ep: 298 | Ep_r: 617.8\n",
"Ep: 299 | Ep_r: 244.01\n",
"Ep: 300 | Ep_r: 687.91\n",
"Ep: 301 | Ep_r: 618.51\n",
"Ep: 302 | Ep_r: 1405.07\n",
"Ep: 303 | Ep_r: 456.95\n",
"Ep: 304 | Ep_r: 340.33\n",
"Ep: 305 | Ep_r: 502.91\n",
"Ep: 306 | Ep_r: 441.21\n",
"Ep: 307 | Ep_r: 255.81\n",
"Ep: 308 | Ep_r: 403.03\n",
"Ep: 309 | Ep_r: 229.1\n",
"Ep: 310 | Ep_r: 308.49\n",
"Ep: 311 | Ep_r: 165.37\n",
"Ep: 312 | Ep_r: 153.76\n",
"Ep: 313 | Ep_r: 442.05\n",
"Ep: 314 | Ep_r: 229.23\n",
"Ep: 315 | Ep_r: 128.52\n",
"Ep: 316 | Ep_r: 358.18\n",
"Ep: 317 | Ep_r: 319.03\n",
"Ep: 318 | Ep_r: 381.76\n",
"Ep: 319 | Ep_r: 199.19\n",
"Ep: 320 | Ep_r: 418.63\n",
"Ep: 321 | Ep_r: 223.95\n",
"Ep: 322 | Ep_r: 222.37\n",
"Ep: 323 | Ep_r: 405.4\n",
"Ep: 324 | Ep_r: 311.32\n",
"Ep: 325 | Ep_r: 184.85\n",
"Ep: 326 | Ep_r: 1026.71\n",
"Ep: 327 | Ep_r: 252.41\n",
"Ep: 328 | Ep_r: 224.93\n",
"Ep: 329 | Ep_r: 620.02\n",
"Ep: 330 | Ep_r: 174.54\n",
"Ep: 331 | Ep_r: 782.45\n",
"Ep: 332 | Ep_r: 263.79\n",
"Ep: 333 | Ep_r: 178.63\n",
"Ep: 334 | Ep_r: 242.84\n",
"Ep: 335 | Ep_r: 635.43\n",
"Ep: 336 | Ep_r: 668.89\n",
"Ep: 337 | Ep_r: 265.42\n",
"Ep: 338 | Ep_r: 207.81\n",
"Ep: 339 | Ep_r: 293.09\n",
"Ep: 340 | Ep_r: 530.23\n",
"Ep: 341 | Ep_r: 479.26\n",
"Ep: 342 | Ep_r: 559.77\n",
"Ep: 343 | Ep_r: 241.39\n",
"Ep: 344 | Ep_r: 158.83\n",
"Ep: 345 | Ep_r: 1510.69\n",
"Ep: 346 | Ep_r: 425.17\n",
"Ep: 347 | Ep_r: 266.94\n",
"Ep: 348 | Ep_r: 166.08\n",
"Ep: 349 | Ep_r: 630.52\n",
"Ep: 350 | Ep_r: 250.95\n",
"Ep: 351 | Ep_r: 625.88\n",
"Ep: 352 | Ep_r: 417.7\n",
"Ep: 353 | Ep_r: 867.81\n",
"Ep: 354 | Ep_r: 150.62\n",
"Ep: 355 | Ep_r: 230.89\n",
"Ep: 356 | Ep_r: 1017.52\n",
"Ep: 357 | Ep_r: 190.28\n",
"Ep: 358 | Ep_r: 396.91\n",
"Ep: 359 | Ep_r: 305.53\n",
"Ep: 360 | Ep_r: 131.61\n",
"Ep: 361 | Ep_r: 387.54\n",
"Ep: 362 | Ep_r: 298.82\n",
"Ep: 363 | Ep_r: 207.56\n",
"Ep: 364 | Ep_r: 248.56\n",
"Ep: 365 | Ep_r: 589.12\n",
"Ep: 366 | Ep_r: 179.52\n",
"Ep: 367 | Ep_r: 130.19\n",
"Ep: 368 | Ep_r: 1220.84\n",
"Ep: 369 | Ep_r: 126.35\n",
"Ep: 370 | Ep_r: 133.31\n",
"Ep: 371 | Ep_r: 485.81\n",
"Ep: 372 | Ep_r: 823.4\n",
"Ep: 373 | Ep_r: 253.26\n",
"Ep: 374 | Ep_r: 466.06\n",
"Ep: 375 | Ep_r: 203.27\n",
"Ep: 376 | Ep_r: 386.5\n",
"Ep: 377 | Ep_r: 491.02\n",
"Ep: 378 | Ep_r: 239.45\n",
"Ep: 379 | Ep_r: 276.93\n",
"Ep: 380 | Ep_r: 331.98\n",
"Ep: 381 | Ep_r: 764.79\n",
"Ep: 382 | Ep_r: 198.29\n",
"Ep: 383 | Ep_r: 717.18\n",
"Ep: 384 | Ep_r: 562.15\n",
"Ep: 385 | Ep_r: 29.44\n",
"Ep: 386 | Ep_r: 344.95\n",
"Ep: 387 | Ep_r: 671.87\n",
"Ep: 388 | Ep_r: 299.81\n",
"Ep: 389 | Ep_r: 899.76\n",
"Ep: 390 | Ep_r: 319.04\n",
"Ep: 391 | Ep_r: 252.11\n",
"Ep: 392 | Ep_r: 865.62\n",
"Ep: 393 | Ep_r: 255.64\n",
"Ep: 394 | Ep_r: 81.74\n",
"Ep: 395 | Ep_r: 213.13\n",
"Ep: 396 | Ep_r: 422.33\n",
"Ep: 397 | Ep_r: 167.47\n",
"Ep: 398 | Ep_r: 507.34\n",
"Ep: 399 | Ep_r: 614.0\n"
]
}
],
"source": [
"\n",
"print('\\nCollecting experience...')\n",
"for i_episode in range(400):\n",
" s = env.reset()\n",
" ep_r = 0\n",
" while True:\n",
" env.render()\n",
" a = dqn.choose_action(s)\n",
"\n",
" # take action\n",
" s_, r, done, info = env.step(a)\n",
"\n",
" # modify the reward\n",
" x, x_dot, theta, theta_dot = s_\n",
" r1 = (env.x_threshold - abs(x)) / env.x_threshold - 0.8\n",
" r2 = (env.theta_threshold_radians - abs(theta)) / env.theta_threshold_radians - 0.5\n",
" r = r1 + r2\n",
"\n",
" dqn.store_transition(s, a, r, s_)\n",
"\n",
" ep_r += r\n",
" if dqn.memory_counter > MEMORY_CAPACITY:\n",
" dqn.learn()\n",
" if done:\n",
" print('Ep: ', i_episode,\n",
" '| Ep_r: ', round(ep_r, 2))\n",
"\n",
" if done:\n",
" break\n",
" s = s_"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.5.2"
}
},
"nbformat": 4,
"nbformat_minor": 2
}

286
tutorial-contents-notebooks/406_GAN.ipynb Normal file
View File

File diff suppressed because one or more lines are too long

262
tutorial-contents-notebooks/501_why_torch_dynamic_graph.ipynb Normal file
View File

File diff suppressed because one or more lines are too long

330
tutorial-contents-notebooks/502_GPU.ipynb Normal file
View File

File diff suppressed because one or more lines are too long

295
tutorial-contents-notebooks/503_dropout.ipynb Normal file
View File

File diff suppressed because one or more lines are too long

371
tutorial-contents-notebooks/504_batch_normalization.ipynb Normal file
View File

File diff suppressed because one or more lines are too long