Skip to content

GitLab

Switch branch/tag
Find file Normal viewHistoryPermalink
Retiarii_example_one-shot_NAS.ipynb 21.3 KB
Newer Older
qianyj's avatar
create branch for v2.9  
qianyj committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354 
{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Retiarii Example - One-shot NAS"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This example will show Retiarii's ability to **express** and **explore** the model space for Neural Architecture Search and Hyper-Parameter Tuning in a simple way. The video demo is in [YouTube](https://youtu.be/3nEx9GMHYEk) and [Bilibili](https://www.bilibili.com/video/BV1c54y1V7vx/).\n",
    "\n",
    "Let's start the journey with Retiarii!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 1: Express the Model Space"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Model space is defined by users to express a set of models that they want to explore, which contains potentially good-performing models. In Retiarii framework, a model space is defined with two parts: a base model and possible mutations on the base model."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Step 1.1: Define the Base Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Defining a base model is almost the same as defining a PyTorch (or TensorFlow) model. Usually, you only need to replace the code ``import torch.nn as nn`` with ``import nni.retiarii.nn.pytorch as nn`` to use NNI wrapped PyTorch modules. Below is a very simple example of defining a base model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch.nn.functional as F\n",
    "import nni.retiarii.nn.pytorch as nn\n",
    "\n",
    "class Net(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(Net, self).__init__()\n",
    "        self.conv1 = nn.Conv2d(3, 6, 3, padding=1)\n",
    "        self.pool = nn.MaxPool2d(2, 2)\n",
    "        self.conv2 = nn.Conv2d(6, 16, 3, padding=1)\n",
    "        self.conv3 = nn.Conv2d(16, 16, 1)\n",
    "\n",
    "        self.bn = nn.BatchNorm2d(16)\n",
    "\n",
    "        self.gap = nn.AdaptiveAvgPool2d(4)\n",
    "        self.fc1 = nn.Linear(16 * 4 * 4, 120)\n",
    "        self.fc2 = nn.Linear(120, 84)\n",
    "        self.fc3 = nn.Linear(84, 10)\n",
    "\n",
    "    def forward(self, x):\n",
    "        bs = x.size(0)\n",
    "\n",
    "        x = self.pool(F.relu(self.conv1(x)))\n",
    "        x0 = F.relu(self.conv2(x))\n",
    "        x1 = F.relu(self.conv3(x0))\n",
    "\n",
    "        x1 += x0\n",
    "        x = self.pool(self.bn(x1))\n",
    "\n",
    "        x = self.gap(x).view(bs, -1)\n",
    "        x = F.relu(self.fc1(x))\n",
    "        x = F.relu(self.fc2(x))\n",
    "        x = self.fc3(x)\n",
    "        return x\n",
    "    \n",
    "model = Net()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Step 1.2: Define the Model Mutations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A base model is only one concrete model, not a model space. NNI provides APIs and primitives for users to express how the base model can be mutated, i.e., a model space that includes many models. The following will use inline Mutation APIs ``LayerChoice`` to choose a layer from candidate operations and use ``InputChoice`` to try out skip connection."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch.nn.functional as F\n",
    "import nni.retiarii.nn.pytorch as nn\n",
    "\n",
    "class Net(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(Net, self).__init__()\n",
    "        # self.conv1 = nn.Conv2d(3, 6, 3, padding=1)\n",
    "        self.conv1 = nn.LayerChoice([nn.Conv2d(3, 6, 3, padding=1), nn.Conv2d(3, 6, 5, padding=2)])\n",
    "        self.pool = nn.MaxPool2d(2, 2)\n",
    "        # self.conv2 = nn.Conv2d(6, 16, 3, padding=1)\n",
    "        self.conv2 = nn.LayerChoice([nn.Conv2d(6, 16, 3, padding=1), nn.Conv2d(6, 16, 5, padding=2)])\n",
    "        self.conv3 = nn.Conv2d(16, 16, 1)\n",
    "\n",
    "        self.skipconnect = nn.InputChoice(n_candidates=2)\n",
    "        self.bn = nn.BatchNorm2d(16)\n",
    "\n",
    "        self.gap = nn.AdaptiveAvgPool2d(4)\n",
    "        self.fc1 = nn.Linear(16 * 4 * 4, 120)\n",
    "        self.fc2 = nn.Linear(120, 84)\n",
    "        self.fc3 = nn.Linear(84, 10)\n",
    "\n",
    "    def forward(self, x):\n",
    "        bs = x.size(0)\n",
    "\n",
    "        x = self.pool(F.relu(self.conv1(x)))\n",
    "        x0 = F.relu(self.conv2(x))\n",
    "        x1 = F.relu(self.conv3(x0))\n",
    "\n",
    "        x1 = self.skipconnect([x1, x1+x0])\n",
    "        x = self.pool(self.bn(x1))\n",
    "\n",
    "        x = self.gap(x).view(bs, -1)\n",
    "        x = F.relu(self.fc1(x))\n",
    "        x = F.relu(self.fc2(x))\n",
    "        x = self.fc3(x)\n",
    "        return x\n",
    "    \n",
    "model = Net()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 2: Explore the Model Space"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "With a defined model space, users can explore the space in two ways. One is the multi-trial NAS method, which searchs by evaluating each sampled model independently. The other is using one-shot weight-sharing based search, which consumes much less computational resource compared to the first one. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this part, we focus on this **one-shot** approach. The principle of the One-shot approach is combining all the models in a model space into one big model (usually called super-model or super-graph). It takes charge of both search, training and testing, by training and evaluating this big model.\n",
    "\n",
    "Retiarii has supported some classic one-shot trainers, like DARTS trainer, ENAS trainer, ProxylessNAS trainer, Single-path trainer, and users can customize a new one-shot trainer according to the APIs provided by Retiarii conveniently.\n",
    "\n",
    "Here, we show an example to use DARTS trainer manually."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Files already downloaded and verified\n",
      "[2021-06-07 11:12:22] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [1/391]  acc1 0.093750 (0.093750)  loss 2.286068 (2.286068)\n",
      "[2021-06-07 11:12:22] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [11/391]  acc1 0.093750 (0.089489)  loss 2.328799 (2.309416)\n",
      "[2021-06-07 11:12:23] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [21/391]  acc1 0.093750 (0.092262)  loss 2.302527 (2.309082)\n",
      "[2021-06-07 11:12:23] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [31/391]  acc1 0.109375 (0.099294)  loss 2.294730 (2.304962)\n",
      "[2021-06-07 11:12:23] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [41/391]  acc1 0.203125 (0.103277)  loss 2.284227 (2.302716)\n",
      "[2021-06-07 11:12:23] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [51/391]  acc1 0.078125 (0.106618)  loss 2.308704 (2.300639)\n",
      "[2021-06-07 11:12:23] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [61/391]  acc1 0.203125 (0.110143)  loss 2.258595 (2.298042)\n",
      "[2021-06-07 11:12:23] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [71/391]  acc1 0.078125 (0.112896)  loss 2.276706 (2.294709)\n",
      "[2021-06-07 11:12:24] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [81/391]  acc1 0.078125 (0.116898)  loss 2.309119 (2.292235)\n",
      "[2021-06-07 11:12:24] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [91/391]  acc1 0.093750 (0.118304)  loss 2.263757 (2.289659)\n",
      "[2021-06-07 11:12:24] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [101/391]  acc1 0.109375 (0.119431)  loss 2.260739 (2.287132)\n",
      "[2021-06-07 11:12:24] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [111/391]  acc1 0.109375 (0.121481)  loss 2.279930 (2.284314)\n",
      "[2021-06-07 11:12:24] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [121/391]  acc1 0.046875 (0.122934)  loss 2.270205 (2.281701)\n",
      "[2021-06-07 11:12:25] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [131/391]  acc1 0.156250 (0.125477)  loss 2.270163 (2.278612)\n",
      "[2021-06-07 11:12:25] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [141/391]  acc1 0.171875 (0.126551)  loss 2.233467 (2.276326)\n",
      "[2021-06-07 11:12:25] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [151/391]  acc1 0.109375 (0.127897)  loss 2.264694 (2.274296)\n",
      "[2021-06-07 11:12:25] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [161/391]  acc1 0.250000 (0.132279)  loss 2.259590 (2.271723)\n",
      "[2021-06-07 11:12:25] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [171/391]  acc1 0.093750 (0.134868)  loss 2.240986 (2.269037)\n",
      "[2021-06-07 11:12:25] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [181/391]  acc1 0.218750 (0.137690)  loss 2.218153 (2.266567)\n",
      "[2021-06-07 11:12:25] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [191/391]  acc1 0.078125 (0.140134)  loss 2.260816 (2.264373)\n",
      "[2021-06-07 11:12:26] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [201/391]  acc1 0.156250 (0.144123)  loss 2.191213 (2.261285)\n",
      "[2021-06-07 11:12:26] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [211/391]  acc1 0.125000 (0.146919)  loss 2.245425 (2.258747)\n",
      "[2021-06-07 11:12:26] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [221/391]  acc1 0.218750 (0.150028)  loss 2.216708 (2.255553)\n",
      "[2021-06-07 11:12:26] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [231/391]  acc1 0.250000 (0.153003)  loss 2.195549 (2.252894)\n",
      "[2021-06-07 11:12:26] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [241/391]  acc1 0.234375 (0.155666)  loss 2.169693 (2.249465)\n",
      "[2021-06-07 11:12:26] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [251/391]  acc1 0.218750 (0.158989)  loss 2.174878 (2.246355)\n",
      "[2021-06-07 11:12:27] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [261/391]  acc1 0.312500 (0.162775)  loss 2.117693 (2.243113)\n",
      "[2021-06-07 11:12:27] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [271/391]  acc1 0.265625 (0.166686)  loss 2.136203 (2.239288)\n",
      "[2021-06-07 11:12:27] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [281/391]  acc1 0.234375 (0.169095)  loss 2.213463 (2.236377)\n",
      "[2021-06-07 11:12:27] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [291/391]  acc1 0.218750 (0.171338)  loss 2.114096 (2.232892)\n",
      "[2021-06-07 11:12:27] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [301/391]  acc1 0.203125 (0.173432)  loss 2.134074 (2.229637)\n",
      "[2021-06-07 11:12:28] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [311/391]  acc1 0.265625 (0.175291)  loss 2.041354 (2.225920)\n",
      "[2021-06-07 11:12:28] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [321/391]  acc1 0.250000 (0.176840)  loss 2.081122 (2.222280)\n",
      "[2021-06-07 11:12:28] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [331/391]  acc1 0.140625 (0.178578)  loss 2.124206 (2.219168)\n",
      "[2021-06-07 11:12:28] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [341/391]  acc1 0.250000 (0.180169)  loss 2.077291 (2.215540)\n",
      "[2021-06-07 11:12:28] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [351/391]  acc1 0.250000 (0.182381)  loss 2.077531 (2.211650)\n",
      "[2021-06-07 11:12:28] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [361/391]  acc1 0.312500 (0.185033)  loss 2.016619 (2.207455)\n",
      "[2021-06-07 11:12:29] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [371/391]  acc1 0.250000 (0.187163)  loss 2.139604 (2.202785)\n",
      "[2021-06-07 11:12:29] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [381/391]  acc1 0.281250 (0.189099)  loss 2.033739 (2.198564)\n",
      "[2021-06-07 11:12:29] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [1/2] Step [391/391]  acc1 0.275000 (0.190441)  loss 1.988353 (2.194509)\n",
      "[2021-06-07 11:12:29] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [1/391]  acc1 0.296875 (0.296875)  loss 2.083627 (2.083627)\n",
      "[2021-06-07 11:12:30] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [11/391]  acc1 0.265625 (0.251420)  loss 2.042856 (2.050898)\n",
      "[2021-06-07 11:12:30] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [21/391]  acc1 0.234375 (0.273065)  loss 2.005307 (2.021047)\n",
      "[2021-06-07 11:12:30] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [31/391]  acc1 0.375000 (0.269657)  loss 1.934093 (2.014375)\n",
      "[2021-06-07 11:12:30] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [41/391]  acc1 0.265625 (0.277439)  loss 2.007705 (2.003260)\n",
      "[2021-06-07 11:12:30] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [51/391]  acc1 0.218750 (0.278799)  loss 2.014602 (2.001039)\n",
      "[2021-06-07 11:12:31] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [61/391]  acc1 0.187500 (0.278945)  loss 2.088407 (1.995837)\n",
      "[2021-06-07 11:12:31] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [71/391]  acc1 0.343750 (0.285651)  loss 1.894479 (1.988130)\n",
      "[2021-06-07 11:12:31] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [81/391]  acc1 0.281250 (0.289159)  loss 1.869002 (1.979012)\n",
      "[2021-06-07 11:12:31] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [91/391]  acc1 0.265625 (0.291552)  loss 1.848354 (1.971483)\n",
      "[2021-06-07 11:12:31] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [101/391]  acc1 0.406250 (0.290996)  loss 1.840711 (1.964297)\n",
      "[2021-06-07 11:12:31] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [111/391]  acc1 0.390625 (0.294764)  loss 1.905811 (1.958954)\n",
      "[2021-06-07 11:12:32] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [121/391]  acc1 0.250000 (0.296617)  loss 1.935214 (1.952315)\n",
      "[2021-06-07 11:12:32] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [131/391]  acc1 0.281250 (0.299618)  loss 1.901846 (1.944634)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2021-06-07 11:12:32] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [141/391]  acc1 0.312500 (0.302970)  loss 1.854658 (1.939751)\n",
      "[2021-06-07 11:12:32] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [151/391]  acc1 0.218750 (0.305257)  loss 1.927818 (1.934704)\n",
      "[2021-06-07 11:12:32] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [161/391]  acc1 0.343750 (0.307648)  loss 1.820810 (1.927533)\n",
      "[2021-06-07 11:12:33] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [171/391]  acc1 0.312500 (0.307383)  loss 1.800313 (1.924665)\n",
      "[2021-06-07 11:12:33] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [181/391]  acc1 0.484375 (0.307925)  loss 1.637479 (1.920402)\n",
      "[2021-06-07 11:12:33] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [191/391]  acc1 0.359375 (0.306692)  loss 1.732374 (1.917680)\n",
      "[2021-06-07 11:12:33] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [201/391]  acc1 0.406250 (0.309624)  loss 1.870701 (1.911484)\n",
      "[2021-06-07 11:12:33] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [211/391]  acc1 0.328125 (0.311982)  loss 1.785704 (1.905039)\n",
      "[2021-06-07 11:12:33] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [221/391]  acc1 0.265625 (0.312712)  loss 1.738683 (1.901547)\n",
      "[2021-06-07 11:12:33] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [231/391]  acc1 0.359375 (0.315409)  loss 1.827117 (1.894860)\n",
      "[2021-06-07 11:12:34] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [241/391]  acc1 0.375000 (0.317881)  loss 1.717454 (1.888916)\n",
      "[2021-06-07 11:12:34] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [251/391]  acc1 0.328125 (0.318663)  loss 1.873310 (1.886883)\n",
      "[2021-06-07 11:12:34] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [261/391]  acc1 0.390625 (0.320163)  loss 1.657088 (1.881767)\n",
      "[2021-06-07 11:12:34] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [271/391]  acc1 0.421875 (0.321264)  loss 1.710897 (1.877521)\n",
      "[2021-06-07 11:12:34] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [281/391]  acc1 0.421875 (0.321230)  loss 1.760745 (1.875136)\n",
      "[2021-06-07 11:12:34] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [291/391]  acc1 0.375000 (0.321413)  loss 1.669255 (1.872129)\n",
      "[2021-06-07 11:12:34] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [301/391]  acc1 0.328125 (0.322051)  loss 1.728873 (1.868047)\n",
      "[2021-06-07 11:12:35] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [311/391]  acc1 0.375000 (0.323000)  loss 1.754761 (1.864783)\n",
      "[2021-06-07 11:12:35] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [321/391]  acc1 0.437500 (0.324864)  loss 1.666240 (1.859164)\n",
      "[2021-06-07 11:12:35] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [331/391]  acc1 0.421875 (0.325954)  loss 1.661471 (1.856318)\n",
      "[2021-06-07 11:12:35] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [341/391]  acc1 0.328125 (0.326475)  loss 1.737106 (1.853075)\n",
      "[2021-06-07 11:12:35] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [351/391]  acc1 0.343750 (0.327724)  loss 1.789253 (1.849491)\n",
      "[2021-06-07 11:12:36] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [361/391]  acc1 0.250000 (0.328558)  loss 1.773805 (1.846033)\n",
      "[2021-06-07 11:12:36] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [371/391]  acc1 0.312500 (0.329094)  loss 1.901358 (1.844091)\n",
      "[2021-06-07 11:12:36] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [381/391]  acc1 0.250000 (0.330011)  loss 1.863921 (1.841390)\n",
      "[2021-06-07 11:12:36] INFO (nni.retiarii.oneshot.pytorch.darts/MainThread) Epoch [2/2] Step [391/391]  acc1 0.325000 (0.331514)  loss 1.729926 (1.837162)\n"
     ]
    }
   ],
   "source": [
    "import torch\n",
    "from utils import accuracy\n",
    "from torchvision import transforms\n",
    "from torchvision.datasets import CIFAR10\n",
    "from nni.retiarii.oneshot.pytorch import DartsTrainer\n",
    "\n",
    "criterion = torch.nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.SGD(model.parameters(), lr=0.001, momentum=0.9)\n",
    "\n",
    "transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])\n",
    "train_dataset = CIFAR10(root=\"./data\", train=True, download=True, transform=transform)\n",
    "\n",
    "trainer = DartsTrainer(\n",
    "    model=model,\n",
    "    loss=criterion,\n",
    "    metrics=lambda output, target: accuracy(output, target),\n",
    "    optimizer=optimizer,\n",
    "    num_epochs=2,\n",
    "    dataset=train_dataset,\n",
    "    batch_size=64,\n",
    "    log_frequency=10\n",
    "    )\n",
    "\n",
    "trainer.fit()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Similarly, the optimal structure found can be exported."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Final architecture: {'_mutation_1': 1, '_mutation_2': 1, '_mutation_3': [1]}\n"
     ]
    }
   ],
   "source": [
    "print('Final architecture:', trainer.export())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}