forked from fastai/fastai
-
Notifications
You must be signed in to change notification settings - Fork 0
/
callbacks.lr_finder.html
446 lines (333 loc) · 31.2 KB
/
callbacks.lr_finder.html
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
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
---
title: callbacks.lr_finder
keywords: fastai
sidebar: home_sidebar
summary: "Implementation of the LR Range test from Leslie Smith"
---
<div class="container" id="notebook-container">
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<h1 id="Learning-Rate-Finder">Learning Rate Finder<a class="anchor-link" href="#Learning-Rate-Finder">¶</a></h1>
</div>
</div>
</div>
<div class="cell border-box-sizing code_cell rendered">
</div>
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<p>Learning rate finder plots lr vs loss relationship for a <a href="/basic_train.html#Learner"><code>Learner</code></a>. The idea is to reduce the amount of guesswork on picking a good starting learning rate.</p>
<p><strong>Overview:</strong></p>
<ol>
<li>First run lr_find <code>learn.lr_find()</code></li>
<li>Plot the learning rate vs loss <code>learn.recorder.plot()</code></li>
<li>Pick a learning rate before it diverges then start training</li>
</ol>
<p><strong>Technical Details:</strong> (first <a href="'https://arxiv.org/abs/1506.01186'">described</a> by Leslie Smith)</p>
<blockquote><p>Train <a href="/basic_train.html#Learner"><code>Learner</code></a> over a few iterations. Start with a very low <code>start_lr</code> and change it at each mini-batch until it reaches a very high <code>end_lr</code>. <a href="/basic_train.html#Recorder"><code>Recorder</code></a> will record the loss at each iteration. Plot those losses against the learning rate to find the optimal value before it diverges.</p>
</blockquote>
</div>
</div>
</div>
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<h2 id="Choosing-a-good-learning-rate">Choosing a good learning rate<a class="anchor-link" href="#Choosing-a-good-learning-rate">¶</a></h2>
</div>
</div>
</div>
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<p>For a more intuitive explanation, please check out <a href="https://sgugger.github.io/how-do-you-find-a-good-learning-rate.html">Sylvain Gugger's post</a></p>
</div>
</div>
</div>
<div class="cell border-box-sizing code_cell rendered">
<div class="input">
<div class="inner_cell">
<div class="input_area">
<div class=" highlight hl-ipython3"><pre><span></span><span class="n">path</span> <span class="o">=</span> <span class="n">untar_data</span><span class="p">(</span><span class="n">URLs</span><span class="o">.</span><span class="n">MNIST_SAMPLE</span><span class="p">)</span>
<span class="n">data</span> <span class="o">=</span> <span class="n">ImageDataBunch</span><span class="o">.</span><span class="n">from_folder</span><span class="p">(</span><span class="n">path</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">simple_learner</span><span class="p">():</span> <span class="k">return</span> <span class="n">Learner</span><span class="p">(</span><span class="n">data</span><span class="p">,</span> <span class="n">simple_cnn</span><span class="p">((</span><span class="mi">3</span><span class="p">,</span><span class="mi">16</span><span class="p">,</span><span class="mi">16</span><span class="p">,</span><span class="mi">2</span><span class="p">)),</span> <span class="n">metrics</span><span class="o">=</span><span class="p">[</span><span class="n">accuracy</span><span class="p">])</span>
<span class="n">learn</span> <span class="o">=</span> <span class="n">simple_learner</span><span class="p">()</span>
</pre></div>
</div>
</div>
</div>
</div>
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<p>First we run this command to launch the search:</p>
</div>
</div>
</div>
<div class="cell border-box-sizing code_cell rendered">
<div class="output_wrapper">
<div class="output">
<div class="output_area">
<div class="output_markdown rendered_html output_subarea ">
<h4 id="lr_find"><code>lr_find</code><a href="https://github.com/fastai/fastai/blob/master/fastai/train.py#L24" class="source_link">[source]</a></h4><blockquote><p><code>lr_find</code>(<strong><code>learn</code></strong>:<a href="/basic_train.html#Learner"><code>Learner</code></a>, <strong><code>start_lr</code></strong>:<code>Floats</code>=<strong><em><code>1e-07</code></em></strong>, <strong><code>end_lr</code></strong>:<code>Floats</code>=<strong><em><code>10</code></em></strong>, <strong><code>num_it</code></strong>:<code>int</code>=<strong><em><code>100</code></em></strong>, <strong><code>stop_div</code></strong>:<code>bool</code>=<strong><em><code>True</code></em></strong>, <strong><code>wd</code></strong>:<code>float</code>=<strong><em><code>None</code></em></strong>)</p>
</blockquote>
<p>Explore lr from <code>start_lr</code> to <code>end_lr</code> over <code>num_it</code> iterations in <code>learn</code>. If <code>stop_div</code>, stops when loss diverges.</p>
</div>
</div>
</div>
</div>
</div>
<div class="cell border-box-sizing code_cell rendered">
<div class="input">
<div class="inner_cell">
<div class="input_area">
<div class=" highlight hl-ipython3"><pre><span></span><span class="n">learn</span><span class="o">.</span><span class="n">lr_find</span><span class="p">(</span><span class="n">stop_div</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span> <span class="n">num_it</span><span class="o">=</span><span class="mi">200</span><span class="p">)</span>
</pre></div>
</div>
</div>
</div>
<div class="output_wrapper">
<div class="output">
<div class="output_area">
<div class="output_subarea output_stream output_stdout output_text">
<pre>LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.
</pre>
</div>
</div>
</div>
</div>
</div>
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<p>Then we plot the loss versus the learning rates. We're interested in finding a good order of magnitude of learning rate, so we plot with a log scale.</p>
</div>
</div>
</div>
<div class="cell border-box-sizing code_cell rendered">
<div class="input">
<div class="inner_cell">
<div class="input_area">
<div class=" highlight hl-ipython3"><pre><span></span><span class="n">learn</span><span class="o">.</span><span class="n">recorder</span><span class="o">.</span><span class="n">plot</span><span class="p">()</span>
</pre></div>
</div>
</div>
</div>
<div class="output_wrapper">
<div class="output">
<div class="output_area">
<div class="output_png output_subarea ">
<img src="data:image/png;base64,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
"
>
</div>
</div>
</div>
</div>
</div>
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<p>Then, we choose a value that is approximately in the middle of the sharpest downward slope. In this case, training with 3e-2 looks like it should work well:</p>
</div>
</div>
</div>
<div class="cell border-box-sizing code_cell rendered">
<div class="input">
<div class="inner_cell">
<div class="input_area">
<div class=" highlight hl-ipython3"><pre><span></span><span class="n">simple_learner</span><span class="p">()</span><span class="o">.</span><span class="n">fit</span><span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mf">1e-2</span><span class="p">)</span>
</pre></div>
</div>
</div>
</div>
<div class="output_wrapper">
<div class="output">
<div class="output_area">
<div class="output_html rendered_html output_subarea ">
Total time: 00:05 <p><table style='width:300px; margin-bottom:10px'>
<tr>
<th>epoch</th>
<th>train_loss</th>
<th>valid_loss</th>
<th>accuracy</th>
</tr>
<tr>
<th>1</th>
<th>0.086687</th>
<th>0.107572</th>
<th>0.965162</th>
</tr>
<tr>
<th>2</th>
<th>0.048850</th>
<th>0.033170</th>
<th>0.988714</th>
</tr>
</table>
</div>
</div>
</div>
</div>
</div>
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<p>Don't just pick the minimum value from the plot!</p>
</div>
</div>
</div>
<div class="cell border-box-sizing code_cell rendered">
<div class="input">
<div class="inner_cell">
<div class="input_area">
<div class=" highlight hl-ipython3"><pre><span></span><span class="n">learn</span> <span class="o">=</span> <span class="n">simple_learner</span><span class="p">()</span>
<span class="n">simple_learner</span><span class="p">()</span><span class="o">.</span><span class="n">fit</span><span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mf">1e-0</span><span class="p">)</span>
</pre></div>
</div>
</div>
</div>
<div class="output_wrapper">
<div class="output">
<div class="output_area">
<div class="output_html rendered_html output_subarea ">
Total time: 00:05 <p><table style='width:300px; margin-bottom:10px'>
<tr>
<th>epoch</th>
<th>train_loss</th>
<th>valid_loss</th>
<th>accuracy</th>
</tr>
<tr>
<th>1</th>
<th>1.705883</th>
<th>0.693147</th>
<th>0.495584</th>
</tr>
<tr>
<th>2</th>
<th>0.713259</th>
<th>0.693147</th>
<th>0.495584</th>
</tr>
</table>
</div>
</div>
</div>
</div>
</div>
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<p>Picking a value before the downward slope results in slow training:</p>
</div>
</div>
</div>
<div class="cell border-box-sizing code_cell rendered">
<div class="input">
<div class="inner_cell">
<div class="input_area">
<div class=" highlight hl-ipython3"><pre><span></span><span class="n">learn</span> <span class="o">=</span> <span class="n">simple_learner</span><span class="p">()</span>
<span class="n">simple_learner</span><span class="p">()</span><span class="o">.</span><span class="n">fit</span><span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mf">1e-3</span><span class="p">)</span>
</pre></div>
</div>
</div>
</div>
<div class="output_wrapper">
<div class="output">
<div class="output_area">
<div class="output_html rendered_html output_subarea ">
Total time: 00:05 <p><table style='width:300px; margin-bottom:10px'>
<tr>
<th>epoch</th>
<th>train_loss</th>
<th>valid_loss</th>
<th>accuracy</th>
</tr>
<tr>
<th>1</th>
<th>0.156533</th>
<th>0.132567</th>
<th>0.955839</th>
</tr>
<tr>
<th>2</th>
<th>0.116986</th>
<th>0.106817</th>
<th>0.966143</th>
</tr>
</table>
</div>
</div>
</div>
</div>
</div>
<div class="cell border-box-sizing code_cell rendered">
<div class="output_wrapper">
<div class="output">
<div class="output_area">
<div class="output_markdown rendered_html output_subarea ">
<h2 id="LRFinder"><code>class</code> <code>LRFinder</code><a href="https://github.com/fastai/fastai/blob/master/fastai/callbacks/lr_finder.py#L9" class="source_link">[source]</a></h2><blockquote><p><code>LRFinder</code>(<strong><code>learn</code></strong>:<a href="/basic_train.html#Learner"><code>Learner</code></a>, <strong><code>start_lr</code></strong>:<code>float</code>=<strong><em><code>1e-07</code></em></strong>, <strong><code>end_lr</code></strong>:<code>float</code>=<strong><em><code>10</code></em></strong>, <strong><code>num_it</code></strong>:<code>int</code>=<strong><em><code>100</code></em></strong>, <strong><code>stop_div</code></strong>:<code>bool</code>=<strong><em><code>True</code></em></strong>) :: <a href="/basic_train.html#LearnerCallback"><code>LearnerCallback</code></a></p>
</blockquote>
<p>Causes <code>learn</code> to go on a mock training from <code>start_lr</code> to <code>end_lr</code> for <code>num_it</code> iterations.</p>
</div>
</div>
</div>
</div>
</div>
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<h3 id="Callback-methods">Callback methods<a class="anchor-link" href="#Callback-methods">¶</a></h3>
</div>
</div>
</div>
<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
<div class="text_cell_render border-box-sizing rendered_html">
<p>You don't call these yourself - they're called by fastai's <a href="/callback.html#Callback"><code>Callback</code></a> system automatically to enable the class's functionality.</p>
</div>
</div>
</div>
<div class="cell border-box-sizing code_cell rendered">
<div class="output_wrapper">
<div class="output">
<div class="output_area">
<div class="output_markdown rendered_html output_subarea ">
<h4 id="LRFinder.on_train_begin"><code>on_train_begin</code><a href="https://github.com/fastai/fastai/blob/master/fastai/callbacks/lr_finder.py#L20" class="source_link">[source]</a></h4><blockquote><p><code>on_train_begin</code>(<strong><code>pbar</code></strong>, <strong>**<code>kwargs</code></strong>:<code>Any</code>)</p>
</blockquote>
<p>Initialize optimizer and learner hyperparameters.</p>
</div>
</div>
</div>
</div>
</div>
<div class="cell border-box-sizing code_cell rendered">
<div class="output_wrapper">
<div class="output">
<div class="output_area">
<div class="output_markdown rendered_html output_subarea ">
<h4 id="LRFinder.on_batch_end"><code>on_batch_end</code><a href="https://github.com/fastai/fastai/blob/master/fastai/callbacks/lr_finder.py#L28" class="source_link">[source]</a></h4><blockquote><p><code>on_batch_end</code>(<strong><code>iteration</code></strong>:<code>int</code>, <strong><code>smooth_loss</code></strong>:<code>TensorOrNumber</code>, <strong>**<code>kwargs</code></strong>:<code>Any</code>)</p>
</blockquote>
<p>Determine if loss has runaway and we should stop.</p>
</div>
</div>
</div>
</div>
</div>
<div class="cell border-box-sizing code_cell rendered">
<div class="output_wrapper">
<div class="output">
<div class="output_area">
<div class="output_markdown rendered_html output_subarea ">
<h4 id="LRFinder.on_epoch_end"><code>on_epoch_end</code><a href="https://github.com/fastai/fastai/blob/master/fastai/callbacks/lr_finder.py#L37" class="source_link">[source]</a></h4><blockquote><p><code>on_epoch_end</code>(<strong>**<code>kwargs</code></strong>:<code>Any</code>)</p>
</blockquote>
<p>Tell Learner if we need to stop.</p>
</div>
</div>
</div>
</div>
</div>
<div class="cell border-box-sizing code_cell rendered">
<div class="output_wrapper">
<div class="output">
<div class="output_area">
<div class="output_markdown rendered_html output_subarea ">
<h4 id="LRFinder.on_train_end"><code>on_train_end</code><a href="https://github.com/fastai/fastai/blob/master/fastai/callbacks/lr_finder.py#L41" class="source_link">[source]</a></h4><blockquote><p><code>on_train_end</code>(<strong>**<code>kwargs</code></strong>:<code>Any</code>)</p>
</blockquote>
<p>Cleanup learn model weights disturbed during LRFind exploration.</p>
</div>
</div>
</div>
</div>
</div>
</div>