wassname/pytorch-lightning
1
0
Fork 0
mirror of https://github.com/wassname/pytorch-lightning.git synced 2026-09-09 11:32:07 +08:00
Code Issues Packages Projects Releases Wiki Activity
Files
gh-pages
pytorch-lightning/Trainer/Training Loop/index.html
T

896 lines
30 KiB
HTML
Raw Permalink Blame History

<!doctype html>
<html lang="en" class="no-js">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width,initial-scale=1">
<meta http-equiv="x-ua-compatible" content="ie=edge">
<meta name="description" content="Documentation for PyTorch LightningModule, the researcher version of keras.">
<meta name="lang:clipboard.copy" content="Copy to clipboard">
<meta name="lang:clipboard.copied" content="Copied to clipboard">
<meta name="lang:search.language" content="en">
<meta name="lang:search.pipeline.stopwords" content="True">
<meta name="lang:search.pipeline.trimmer" content="True">
<meta name="lang:search.result.none" content="No matching documents">
<meta name="lang:search.result.one" content="1 matching document">
<meta name="lang:search.result.other" content="# matching documents">
<meta name="lang:search.tokenizer" content="[\s\-]+">
<link rel="shortcut icon" href="../../assets/images/favicon.png">
<meta name="generator" content="mkdocs-1.0.4, mkdocs-material-4.4.0">
<title>Training Loop - PyTorch lightning Documentation</title>
<link rel="stylesheet" href="../../assets/stylesheets/application.0284f74d.css">
<script src="../../assets/javascripts/modernizr.74668098.js"></script>
<link href="https://fonts.gstatic.com" rel="preconnect" crossorigin>
<link rel="stylesheet" href="https://fonts.googleapis.com/css?family=Roboto:300,400,400i,700|Roboto+Mono&display=fallback">
<style>body,input{font-family:"Roboto","Helvetica Neue",Helvetica,Arial,sans-serif}code,kbd,pre{font-family:"Roboto Mono","Courier New",Courier,monospace}</style>
<link rel="stylesheet" href="../../assets/fonts/material-icons.css">
</head>
<body dir="ltr">
<svg class="md-svg">
<defs>
<svg xmlns="http://www.w3.org/2000/svg" width="416" height="448" viewBox="0 0 416 448" id="__github"><path fill="currentColor" d="M160 304q0 10-3.125 20.5t-10.75 19T128 352t-18.125-8.5-10.75-19T96 304t3.125-20.5 10.75-19T128 256t18.125 8.5 10.75 19T160 304zm160 0q0 10-3.125 20.5t-10.75 19T288 352t-18.125-8.5-10.75-19T256 304t3.125-20.5 10.75-19T288 256t18.125 8.5 10.75 19T320 304zm40 0q0-30-17.25-51T296 232q-10.25 0-48.75 5.25Q229.5 240 208 240t-39.25-2.75Q130.75 232 120 232q-29.5 0-46.75 21T56 304q0 22 8 38.375t20.25 25.75 30.5 15 35 7.375 37.25 1.75h42q20.5 0 37.25-1.75t35-7.375 30.5-15 20.25-25.75T360 304zm56-44q0 51.75-15.25 82.75-9.5 19.25-26.375 33.25t-35.25 21.5-42.5 11.875-42.875 5.5T212 416q-19.5 0-35.5-.75t-36.875-3.125-38.125-7.5-34.25-12.875T37 371.5t-21.5-28.75Q0 312 0 260q0-59.25 34-99-6.75-20.5-6.75-42.5 0-29 12.75-54.5 27 0 47.5 9.875t47.25 30.875Q171.5 96 212 96q37 0 70 8 26.25-20.5 46.75-30.25T376 64q12.75 25.5 12.75 54.5 0 21.75-6.75 42 34 40 34 99.5z"/></svg>
</defs>
</svg>
<input class="md-toggle" data-md-toggle="drawer" type="checkbox" id="__drawer" autocomplete="off">
<input class="md-toggle" data-md-toggle="search" type="checkbox" id="__search" autocomplete="off">
<label class="md-overlay" data-md-component="overlay" for="__drawer"></label>
<a href="#accumulated-gradients" tabindex="1" class="md-skip">
Skip to content
</a>
<header class="md-header" data-md-component="header">
<nav class="md-header-nav md-grid">
<div class="md-flex">
<div class="md-flex__cell md-flex__cell--shrink">
<a href="../.." title="PyTorch lightning Documentation" class="md-header-nav__button md-logo">
<i class="md-icon"></i>
</a>
</div>
<div class="md-flex__cell md-flex__cell--shrink">
<label class="md-icon md-icon--menu md-header-nav__button" for="__drawer"></label>
</div>
<div class="md-flex__cell md-flex__cell--stretch">
<div class="md-flex__ellipsis md-header-nav__title" data-md-component="title">
<span class="md-header-nav__topic">
PyTorch lightning Documentation
</span>
<span class="md-header-nav__topic">
Training Loop
</span>
</div>
</div>
<div class="md-flex__cell md-flex__cell--shrink">
<label class="md-icon md-icon--search md-header-nav__button" for="__search"></label>
<div class="md-search" data-md-component="search" role="dialog">
<label class="md-search__overlay" for="__search"></label>
<div class="md-search__inner" role="search">
<form class="md-search__form" name="search">
<input type="text" class="md-search__input" name="query" placeholder="Search" autocapitalize="off" autocorrect="off" autocomplete="off" spellcheck="false" data-md-component="query" data-md-state="active">
<label class="md-icon md-search__icon" for="__search"></label>
<button type="reset" class="md-icon md-search__icon" data-md-component="reset" tabindex="-1">
&#xE5CD;
</button>
</form>
<div class="md-search__output">
<div class="md-search__scrollwrap" data-md-scrollfix>
<div class="md-search-result" data-md-component="result">
<div class="md-search-result__meta">
Type to start searching
</div>
<ol class="md-search-result__list"></ol>
</div>
</div>
</div>
</div>
</div>
</div>
<div class="md-flex__cell md-flex__cell--shrink">
<div class="md-header-nav__source">
<a href="https://github.com/williamFalcon/pytorch-lightning/" title="Go to repository" class="md-source" data-md-source="github">
<div class="md-source__icon">
<svg viewBox="0 0 24 24" width="24" height="24">
<use xlink:href="#__github" width="24" height="24"></use>
</svg>
</div>
<div class="md-source__repository">
williamFalcon/pytorch-lightning
</div>
</a>
</div>
</div>
</div>
</nav>
</header>
<div class="md-container">
<main class="md-main">
<div class="md-main__inner md-grid" data-md-component="container">
<div class="md-sidebar md-sidebar--primary" data-md-component="navigation">
<div class="md-sidebar__scrollwrap">
<div class="md-sidebar__inner">
<nav class="md-nav md-nav--primary" data-md-level="0">
<label class="md-nav__title md-nav__title--site" for="__drawer">
<a href="../.." title="PyTorch lightning Documentation" class="md-nav__button md-logo">
<i class="md-icon"></i>
</a>
PyTorch lightning Documentation
</label>
<div class="md-nav__source">
<a href="https://github.com/williamFalcon/pytorch-lightning/" title="Go to repository" class="md-source" data-md-source="github">
<div class="md-source__icon">
<svg viewBox="0 0 24 24" width="24" height="24">
<use xlink:href="#__github" width="24" height="24"></use>
</svg>
</div>
<div class="md-source__repository">
williamFalcon/pytorch-lightning
</div>
</a>
</div>
<ul class="md-nav__list" data-md-scrollfix>
<li class="md-nav__item">
<a href="../.." title="Home" class="md-nav__link">
Home
</a>
</li>
<li class="md-nav__item md-nav__item--nested">
<input class="md-toggle md-nav__toggle" data-md-toggle="nav-2" type="checkbox" id="nav-2">
<label class="md-nav__link" for="nav-2">
LightningModule
</label>
<nav class="md-nav" data-md-component="collapsible" data-md-level="1">
<label class="md-nav__title" for="nav-2">
LightningModule
</label>
<ul class="md-nav__list" data-md-scrollfix>
<li class="md-nav__item">
<a href="../../LightningModule/RequiredTrainerInterface/" title="Lightning Module interface" class="md-nav__link">
Lightning Module interface
</a>
</li>
<li class="md-nav__item">
<a href="../../LightningModule/methods/" title="Methods" class="md-nav__link">
Methods
</a>
</li>
<li class="md-nav__item">
<a href="../../LightningModule/properties/" title="Properties" class="md-nav__link">
Properties
</a>
</li>
</ul>
</nav>
</li>
<li class="md-nav__item md-nav__item--active md-nav__item--nested">
<input class="md-toggle md-nav__toggle" data-md-toggle="nav-3" type="checkbox" id="nav-3" checked>
<label class="md-nav__link" for="nav-3">
Trainer
</label>
<nav class="md-nav" data-md-component="collapsible" data-md-level="1">
<label class="md-nav__title" for="nav-3">
Trainer
</label>
<ul class="md-nav__list" data-md-scrollfix>
<li class="md-nav__item">
<a href="../" title="Trainer" class="md-nav__link">
Trainer
</a>
</li>
<li class="md-nav__item">
<a href="../Checkpointing/" title="Checkpointing" class="md-nav__link">
Checkpointing
</a>
</li>
<li class="md-nav__item">
<a href="../Distributed training/" title="Distributed training" class="md-nav__link">
Distributed training
</a>
</li>
<li class="md-nav__item">
<a href="../Logging/" title="Logging" class="md-nav__link">
Logging
</a>
</li>
<li class="md-nav__item">
<a href="../SLURM Managed Cluster/" title="SLURM Managed Cluster" class="md-nav__link">
SLURM Managed Cluster
</a>
</li>
<li class="md-nav__item">
<a href="../Testing loop/" title="Testing loop" class="md-nav__link">
Testing loop
</a>
</li>
<li class="md-nav__item md-nav__item--active">
<input class="md-toggle md-nav__toggle" data-md-toggle="toc" type="checkbox" id="__toc">
<label class="md-nav__link md-nav__link--active" for="__toc">
Training Loop
</label>
<a href="./" title="Training Loop" class="md-nav__link md-nav__link--active">
Training Loop
</a>
<nav class="md-nav md-nav--secondary">
<label class="md-nav__title" for="__toc">Table of contents</label>
<ul class="md-nav__list" data-md-scrollfix>
<li class="md-nav__item">
<a href="#accumulated-gradients" title="Accumulated gradients" class="md-nav__link">
Accumulated gradients
</a>
</li>
<li class="md-nav__item">
<a href="#force-training-for-min-or-max-epochs" title="Force training for min or max epochs" class="md-nav__link">
Force training for min or max epochs
</a>
</li>
<li class="md-nav__item">
<a href="#early-stopping" title="Early stopping" class="md-nav__link">
Early stopping
</a>
</li>
<li class="md-nav__item">
<a href="#force-disable-early-stop" title="Force disable early stop" class="md-nav__link">
Force disable early stop
</a>
</li>
<li class="md-nav__item">
<a href="#gradient-clipping" title="Gradient Clipping" class="md-nav__link">
Gradient Clipping
</a>
</li>
<li class="md-nav__item">
<a href="#inspect-gradient-norms" title="Inspect gradient norms" class="md-nav__link">
Inspect gradient norms
</a>
</li>
<li class="md-nav__item">
<a href="#set-how-much-of-the-training-set-to-check" title="Set how much of the training set to check" class="md-nav__link">
Set how much of the training set to check
</a>
</li>
<li class="md-nav__item">
<a href="#packed-sequences-as-inputs" title="Packed sequences as inputs" class="md-nav__link">
Packed sequences as inputs
</a>
</li>
<li class="md-nav__item">
<a href="#truncated-back-propagation-through-time" title="Truncated Back Propagation Through Time" class="md-nav__link">
Truncated Back Propagation Through Time
</a>
</li>
</ul>
</nav>
</li>
<li class="md-nav__item">
<a href="../Validation loop/" title="Validation loop" class="md-nav__link">
Validation loop
</a>
</li>
<li class="md-nav__item">
<a href="../debugging/" title="Debugging" class="md-nav__link">
Debugging
</a>
</li>
<li class="md-nav__item">
<a href="../hooks/" title="Hooks" class="md-nav__link">
Hooks
</a>
</li>
</ul>
</nav>
</li>
<li class="md-nav__item md-nav__item--nested">
<input class="md-toggle md-nav__toggle" data-md-toggle="nav-4" type="checkbox" id="nav-4">
<label class="md-nav__link" for="nav-4">
Examples
</label>
<nav class="md-nav" data-md-component="collapsible" data-md-level="1">
<label class="md-nav__title" for="nav-4">
Examples
</label>
<ul class="md-nav__list" data-md-scrollfix>
<li class="md-nav__item">
<a href="../../examples/Examples/" title="Examples" class="md-nav__link">
Examples
</a>
</li>
</ul>
</nav>
</li>
</ul>
</nav>
</div>
</div>
</div>
<div class="md-sidebar md-sidebar--secondary" data-md-component="toc">
<div class="md-sidebar__scrollwrap">
<div class="md-sidebar__inner">
<nav class="md-nav md-nav--secondary">
<label class="md-nav__title" for="__toc">Table of contents</label>
<ul class="md-nav__list" data-md-scrollfix>
<li class="md-nav__item">
<a href="#accumulated-gradients" title="Accumulated gradients" class="md-nav__link">
Accumulated gradients
</a>
</li>
<li class="md-nav__item">
<a href="#force-training-for-min-or-max-epochs" title="Force training for min or max epochs" class="md-nav__link">
Force training for min or max epochs
</a>
</li>
<li class="md-nav__item">
<a href="#early-stopping" title="Early stopping" class="md-nav__link">
Early stopping
</a>
</li>
<li class="md-nav__item">
<a href="#force-disable-early-stop" title="Force disable early stop" class="md-nav__link">
Force disable early stop
</a>
</li>
<li class="md-nav__item">
<a href="#gradient-clipping" title="Gradient Clipping" class="md-nav__link">
Gradient Clipping
</a>
</li>
<li class="md-nav__item">
<a href="#inspect-gradient-norms" title="Inspect gradient norms" class="md-nav__link">
Inspect gradient norms
</a>
</li>
<li class="md-nav__item">
<a href="#set-how-much-of-the-training-set-to-check" title="Set how much of the training set to check" class="md-nav__link">
Set how much of the training set to check
</a>
</li>
<li class="md-nav__item">
<a href="#packed-sequences-as-inputs" title="Packed sequences as inputs" class="md-nav__link">
Packed sequences as inputs
</a>
</li>
<li class="md-nav__item">
<a href="#truncated-back-propagation-through-time" title="Truncated Back Propagation Through Time" class="md-nav__link">
Truncated Back Propagation Through Time
</a>
</li>
</ul>
</nav>
</div>
</div>
</div>
<div class="md-content">
<article class="md-content__inner md-typeset">
<a href="https://github.com/williamFalcon/pytorch-lightning/edit/master/docs/Trainer/Training Loop.md" title="Edit this page" class="md-icon md-content__icon">&#xE3C9;</a>
<h1>Training Loop</h1>
<p>The lightning training loop handles everything except the actual computations of your model. To decide what will happen in your training loop, define the <a href="https://williamfalcon.github.io/pytorch-lightning/LightningModule/RequiredTrainerInterface/#training_step">training_step function</a>.</p>
<p>Below are all the things lightning automates for you in the training loop.</p>
<hr />
<h4 id="accumulated-gradients">Accumulated gradients</h4>
<p>Accumulated gradients runs K small batches of size N before doing a backwards pass. The effect is a large effective batch size of size KxN.</p>
<table class="codehilitetable"><tr><td class="linenos"><div class="linenodiv"><pre>1
2</pre></div></td><td class="code"><div class="codehilite"><pre><span></span><span class="c1"># DEFAULT (ie: no accumulated grads)</span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">accumulate_grad_batches</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
</pre></div>
</td></tr></table>
<hr />
<h4 id="force-training-for-min-or-max-epochs">Force training for min or max epochs</h4>
<p>It can be useful to force training for a minimum number of epochs or limit to a max number</p>
<table class="codehilitetable"><tr><td class="linenos"><div class="linenodiv"><pre>1
2</pre></div></td><td class="code"><div class="codehilite"><pre><span></span><span class="c1"># DEFAULT</span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">min_nb_epochs</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">max_nb_epochs</span><span class="o">=</span><span class="mi">1000</span><span class="p">)</span>
</pre></div>
</td></tr></table>
<hr />
<h4 id="early-stopping">Early stopping</h4>
<p>The trainer already sets up default early stopping for you.
To modify this behavior, pass in your own EarlyStopping callback.</p>
<table class="codehilitetable"><tr><td class="linenos"><div class="linenodiv"><pre> 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19</pre></div></td><td class="code"><div class="codehilite"><pre><span></span><span class="kn">from</span> <span class="nn">pytorch_lightning.callbacks</span> <span class="kn">import</span> <span class="n">EarlyStopping</span>
<span class="c1"># DEFAULTS used by Trainer</span>
<span class="n">early_stop_callback</span> <span class="o">=</span> <span class="n">EarlyStopping</span><span class="p">(</span>
<span class="n">monitor</span><span class="o">=</span><span class="s1">&#39;val_loss&#39;</span><span class="p">,</span>
<span class="n">min_delta</span><span class="o">=</span><span class="mf">0.00</span><span class="p">,</span>
<span class="n">patience</span><span class="o">=</span><span class="mi">3</span><span class="p">,</span>
<span class="n">verbose</span><span class="o">=</span><span class="bp">False</span><span class="p">,</span>
<span class="n">mode</span><span class="o">=</span><span class="s1">&#39;min&#39;</span>
<span class="p">)</span>
<span class="c1"># without passing anything in, uses the default callback above</span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">()</span>
<span class="c1"># pass in your own to override the default callback</span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">early_stop_callback</span><span class="o">=</span><span class="n">early_stop_callback</span><span class="p">)</span>
<span class="c1"># pass in None to disable it</span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">early_stop_callback</span><span class="o">=</span><span class="bp">None</span><span class="p">)</span>
</pre></div>
</td></tr></table>
<hr />
<h4 id="force-disable-early-stop">Force disable early stop</h4>
<p>To disable early stopping pass None to the early_stop_callback</p>
<table class="codehilitetable"><tr><td class="linenos"><div class="linenodiv"><pre>1
2</pre></div></td><td class="code"><div class="codehilite"><pre><span></span><span class="c1"># DEFAULT</span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">early_stop_callback</span><span class="o">=</span><span class="bp">None</span><span class="p">)</span>
</pre></div>
</td></tr></table>
<hr />
<h4 id="gradient-clipping">Gradient Clipping</h4>
<p>Gradient clipping may be enabled to avoid exploding gradients.
Specifically, this will <a href="https://pytorch.org/docs/stable/nn.html#torch.nn.utils.clip_grad_norm_">clip the gradient norm computed over all model parameters <em>together</em></a>.</p>
<table class="codehilitetable"><tr><td class="linenos"><div class="linenodiv"><pre>1
2
3
4
5</pre></div></td><td class="code"><div class="codehilite"><pre><span></span><span class="c1"># DEFAULT (ie: don&#39;t clip)</span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">gradient_clip_val</span><span class="o">=</span><span class="mi">0</span><span class="p">)</span>
<span class="c1"># clip gradients with norm above 0.5</span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">gradient_clip_val</span><span class="o">=</span><span class="mf">0.5</span><span class="p">)</span>
</pre></div>
</td></tr></table>
<hr />
<h4 id="inspect-gradient-norms">Inspect gradient norms</h4>
<p>Looking at grad norms can help you figure out where training might be going wrong.</p>
<table class="codehilitetable"><tr><td class="linenos"><div class="linenodiv"><pre>1
2
3
4
5</pre></div></td><td class="code"><div class="codehilite"><pre><span></span><span class="c1"># DEFAULT (-1 doesn&#39;t track norms)</span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">track_grad_norm</span><span class="o">=-</span><span class="mi">1</span><span class="p">)</span>
<span class="c1"># track the LP norm (P=2 here)</span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">track_grad_norm</span><span class="o">=</span><span class="mi">2</span><span class="p">)</span>
</pre></div>
</td></tr></table>
<hr />
<h4 id="set-how-much-of-the-training-set-to-check">Set how much of the training set to check</h4>
<p>If you don't want to check 100% of the training set (for debugging or if it's huge), set this flag.</p>
<p>train_percent_check will be overwritten by overfit_pct if <code>overfit_pct &gt; 0</code></p>
<table class="codehilitetable"><tr><td class="linenos"><div class="linenodiv"><pre>1
2
3
4
5</pre></div></td><td class="code"><div class="codehilite"><pre><span></span><span class="c1"># DEFAULT</span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">train_percent_check</span><span class="o">=</span><span class="mf">1.0</span><span class="p">)</span>
<span class="c1"># check 10% only</span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">train_percent_check</span><span class="o">=</span><span class="mf">0.1</span><span class="p">)</span>
</pre></div>
</td></tr></table>
<hr />
<h4 id="packed-sequences-as-inputs">Packed sequences as inputs</h4>
<p>When using PackedSequence, do 2 things:
1. return either a padded tensor in dataset or a list of variable length tensors in the dataloader collate_fn (example above shows the list implementation). <br />
2. Pack the sequence in forward or training and validation steps depending on use case.</p>
<table class="codehilitetable"><tr><td class="linenos"><div class="linenodiv"><pre> 1
2
3
4
5
6
7
8
9
10</pre></div></td><td class="code"><div class="codehilite"><pre><span></span><span class="c1"># For use in dataloader</span>
<span class="k">def</span> <span class="nf">collate_fn</span><span class="p">(</span><span class="n">batch</span><span class="p">):</span>
<span class="n">x</span> <span class="o">=</span> <span class="p">[</span><span class="n">item</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span> <span class="k">for</span> <span class="n">item</span> <span class="ow">in</span> <span class="n">batch</span><span class="p">]</span>
<span class="n">y</span> <span class="o">=</span> <span class="p">[</span><span class="n">item</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span> <span class="k">for</span> <span class="n">item</span> <span class="ow">in</span> <span class="n">batch</span><span class="p">]</span>
<span class="k">return</span> <span class="n">x</span><span class="p">,</span> <span class="n">y</span>
<span class="c1"># In module</span>
<span class="k">def</span> <span class="nf">training_step</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">batch</span><span class="p">,</span> <span class="n">batch_nb</span><span class="p">):</span>
<span class="n">x</span> <span class="o">=</span> <span class="n">rnn</span><span class="o">.</span><span class="n">pack_sequence</span><span class="p">(</span><span class="n">batch</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="n">enforce_sorted</span><span class="o">=</span><span class="bp">False</span><span class="p">)</span>
<span class="n">y</span> <span class="o">=</span> <span class="n">rnn</span><span class="o">.</span><span class="n">pack_sequence</span><span class="p">(</span><span class="n">batch</span><span class="p">[</span><span class="mi">1</span><span class="p">],</span> <span class="n">enforce_sorted</span><span class="o">=</span><span class="bp">False</span><span class="p">)</span>
</pre></div>
</td></tr></table>
<hr />
<h4 id="truncated-back-propagation-through-time">Truncated Back Propagation Through Time</h4>
<p>There are times when multiple backwards passes are needed for each batch. For example, it may save memory to use Truncated Back Propagation Through Time when training RNNs.</p>
<p>When this flag is enabled each batch is split into sequences of size truncated_bptt_steps and passed to training_step(...) separately. A default splitting function is provided, however, you can override it for more flexibility. See <a href="https://williamfalcon.github.io/pytorch-lightning/Trainer/hooks#tbptt_split_batch">tbptt_split_batch</a>.</p>
<table class="codehilitetable"><tr><td class="linenos"><div class="linenodiv"><pre>1
2
3
4
5</pre></div></td><td class="code"><div class="codehilite"><pre><span></span><span class="c1"># DEFAULT (single backwards pass per batch)</span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">truncated_bptt_steps</span><span class="o">=</span><span class="bp">None</span><span class="p">)</span>
<span class="c1"># (split batch into sequences of size 2)</span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">truncated_bptt_steps</span><span class="o">=</span><span class="mi">2</span><span class="p">)</span>
</pre></div>
</td></tr></table>
</article>
</div>
</div>
</main>
<footer class="md-footer">
<div class="md-footer-nav">
<nav class="md-footer-nav__inner md-grid">
<a href="../Testing loop/" title="Testing loop" class="md-flex md-footer-nav__link md-footer-nav__link--prev" rel="prev">
<div class="md-flex__cell md-flex__cell--shrink">
<i class="md-icon md-icon--arrow-back md-footer-nav__button"></i>
</div>
<div class="md-flex__cell md-flex__cell--stretch md-footer-nav__title">
<span class="md-flex__ellipsis">
<span class="md-footer-nav__direction">
Previous
</span>
Testing loop
</span>
</div>
</a>
<a href="../Validation loop/" title="Validation loop" class="md-flex md-footer-nav__link md-footer-nav__link--next" rel="next">
<div class="md-flex__cell md-flex__cell--stretch md-footer-nav__title">
<span class="md-flex__ellipsis">
<span class="md-footer-nav__direction">
Next
</span>
Validation loop
</span>
</div>
<div class="md-flex__cell md-flex__cell--shrink">
<i class="md-icon md-icon--arrow-forward md-footer-nav__button"></i>
</div>
</a>
</nav>
</div>
<div class="md-footer-meta md-typeset">
<div class="md-footer-meta__inner md-grid">
<div class="md-footer-copyright">
powered by
<a href="https://www.mkdocs.org">MkDocs</a>
and
<a href="https://squidfunk.github.io/mkdocs-material/">
Material for MkDocs</a>
</div>
</div>
</div>
</footer>
</div>
<script src="../../assets/javascripts/application.245445c6.js"></script>
<script>app.initialize({version:"1.0.4",url:{base:"../.."}})</script>
</body>
</html>
Reference in New Issue View Git Blame Copy Permalink