wassname/pytorch-lightning
1
0
Fork 0
mirror of https://github.com/wassname/pytorch-lightning.git synced 2026-09-10 12:21:57 +08:00
Code Issues Packages Projects Releases Wiki Activity

Deployed f0af138 with MkDocs version: 1.0.4

Browse Source
This commit is contained in:
William Falcon
2019-08-13 15:21:24 -05:00
parent d32c548cf8
commit 9b8a06bfdb
15 changed files with 1240 additions and 598 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Trainer/debugging/index.html
+30 -16
View File
@@ -607,29 +607,41 @@
<h4 id="fast-dev-run">Fast dev run</h4>
<p>This flag is meant for debugging a full train/val/test loop. It'll activate callbacks, everything but only with 1 training and 1 validation batch.
Use this to debug a full run of your program quickly</p>
<pre><code class="python"># DEFAULT
trainer = Trainer(fast_dev_run=False)
</code></pre>
<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">fast_dev_run</span><span class="o">=</span><span class="bp">False</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>
<pre><code class="python"># DEFAULT (-1 doesn't track norms)
trainer = Trainer(track_grad_norm=-1)
<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>
# track the LP norm (P=2 here)
trainer = Trainer(track_grad_norm=2)
</code></pre>
<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="make-model-overfit-on-subset-of-data">Make model overfit on subset of data</h4>
<p>A useful debugging trick is to make your model overfit a tiny fraction of the data.</p>
<pre><code class="python"># DEFAULT don't overfit (ie: normal training)
trainer = Trainer(overfit_pct=0.0)
<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 don&#39;t overfit (ie: normal training)</span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">overfit_pct</span><span class="o">=</span><span class="mf">0.0</span><span class="p">)</span>
# overfit on 1% of data
trainer = Trainer(overfit_pct=0.01)
</code></pre>
<span class="c1"># overfit on 1% of data </span>
<span class="n">trainer</span> <span class="o">=</span> <span class="n">Trainer</span><span class="p">(</span><span class="n">overfit_pct</span><span class="o">=</span><span class="mf">0.01</span><span class="p">)</span>
</pre></div>
</td></tr></table>
<hr />
<h4 id="print-the-parameter-count-by-layer">Print the parameter count by layer</h4>
@@ -637,9 +649,11 @@ trainer = Trainer(overfit_pct=0.01)
<hr />
<h4 id="print-which-gradients-are-nan">Print which gradients are nan</h4>
<p>This option prints a list of tensors with nan gradients.</p>
<pre><code class="python"># DEFAULT
trainer = Trainer(print_nan_grads=False)
</code></pre>
<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">print_nan_grads</span><span class="o">=</span><span class="bp">False</span><span class="p">)</span>
</pre></div>
</td></tr></table>
<hr />
<h4 id="log-gpu-usage">Log GPU usage</h4>