update requirements and documentation

README.md

@@ -1,24 +1,45 @@
-# Project name here
-> Summary description here.
+# dl_pipeline
+> A general deep learning pipeline (in construction) for kaggle competitions and other projects.
 
 
-This file will become your README and also the index of your documentation.
-
 ## Install
 
-`pip install your_project_name`
-
-## How to use
+Setup with pip:
 
-Fill me in please! Don't forget code examples:
+`pip install dl_pipeline`
 
+Clone and editable setup:
+```bash
+git clone https://github.com/mnpinto/dl_pipeline
+cd dl_pipeline
+pip install -e .
 ```
-1+1
-```
-
-
-
-
-    2
-
 
+### Rainforest Connection Species Audio Detection
+
+```bash
+#!/bin/bash
+arch='densenet121'
+model_name='model_0'
+sample_rate=32000
+n_mels=128
+hop_length=640
+
+for fold in 0 1 2 3 4
+do
+    echo "Training $model for fold $fold"
+    kaggle_rainforest2021 --fold $fold --model_name $model_name \
+        --model $arch --sample_rate $sample_rate --n_mels $n_mels \
+        --hop_length $hop_length --bs 32 --head_ps 0.8 \
+        --tile_width 1024 --mixup true >> log.train
+done
+
+for tw in 64 128 256
+do
+    echo "Generate predictions for $model with tile_width of $tw"
+    kaggle_rainforest2021 --run_test true --model_name $model_name \
+        --model $arch --sample_rate $sample_rate --n_mels $n_mels \
+        --hop_length $hop_length --tile_width $tw \
+        --save_preds true >> log.predict
+done
+```

docs/_data/sidebars/home_sidebar.yml

@@ -12,15 +12,21 @@ entries:
     - output: web,pdf
       title: Audio Core
       url: core.html
+    - output: web,pdf
+      title: Vision Losses
+      url: losses.html
+    - output: web,pdf
+      title: Models
+      url: models.html
+    - output: web,pdf
+      title: Triplet Learning
+      url: triplet.html
     - output: web,pdf
       title: Audio Augmentations
       url: augmentations.html
     - output: web,pdf
       title: Audio Datasets
       url: dataset.html
-    - output: web,pdf
-      title: Audio Models
-      url: models.html
     - output: web,pdf
       title: 'Audio Utility Functions '
       url: util.html

docs/core.html

@@ -54,7 +54,7 @@
 
 
 <div class="output_markdown rendered_html output_subarea ">
-<h2 id="TensorAudio" class="doc_header"><code>class</code> <code>TensorAudio</code><a href="https://github.com/mnpinto/dl_pipeline/tree/master/dl_pipeline/audio/core.py#L17" class="source_link" style="float:right">[source]</a></h2><blockquote><p><code>TensorAudio</code>(<strong><code>x</code></strong>, <strong>**<code>kwargs</code></strong>) :: <code>TensorBase</code></p>
+<h2 id="TensorAudio" class="doc_header"><code>class</code> <code>TensorAudio</code><a href="https://github.com/mnpinto/dl_pipeline/tree/master/dl_pipeline/audio/core.py#L20" class="source_link" style="float:right">[source]</a></h2><blockquote><p><code>TensorAudio</code>(<strong><code>x</code></strong>, <strong>**<code>kwargs</code></strong>) :: <code>TensorBase</code></p>
 </blockquote>
 
 </div>
@@ -78,7 +78,7 @@ <h2 id="TensorAudio" class="doc_header"><code>class</code> <code>TensorAudio</co
 
 
 <div class="output_markdown rendered_html output_subarea ">
-<h2 id="TensorAudioLabel" class="doc_header"><code>class</code> <code>TensorAudioLabel</code><a href="https://github.com/mnpinto/dl_pipeline/tree/master/dl_pipeline/audio/core.py#L18" class="source_link" style="float:right">[source]</a></h2><blockquote><p><code>TensorAudioLabel</code>(<strong><code>x</code></strong>, <strong>**<code>kwargs</code></strong>) :: <code>TensorBase</code></p>
+<h2 id="TensorAudioLabel" class="doc_header"><code>class</code> <code>TensorAudioLabel</code><a href="https://github.com/mnpinto/dl_pipeline/tree/master/dl_pipeline/audio/core.py#L21" class="source_link" style="float:right">[source]</a></h2><blockquote><p><code>TensorAudioLabel</code>(<strong><code>x</code></strong>, <strong>**<code>kwargs</code></strong>) :: <code>TensorBase</code></p>
 </blockquote>
 
 </div>
@@ -102,7 +102,7 @@ <h2 id="TensorAudioLabel" class="doc_header"><code>class</code> <code>TensorAudi
 
 
 <div class="output_markdown rendered_html output_subarea ">
-<h4 id="load_npy" class="doc_header"><code>load_npy</code><a href="https://github.com/mnpinto/dl_pipeline/tree/master/dl_pipeline/audio/core.py#L20" class="source_link" style="float:right">[source]</a></h4><blockquote><p><code>load_npy</code>(<strong><code>fn</code></strong>)</p>
+<h4 id="load_npy" class="doc_header"><code>load_npy</code><a href="https://github.com/mnpinto/dl_pipeline/tree/master/dl_pipeline/audio/core.py#L23" class="source_link" style="float:right">[source]</a></h4><blockquote><p><code>load_npy</code>(<strong><code>fn</code></strong>)</p>
 </blockquote>
 
 </div>
@@ -126,7 +126,7 @@ <h4 id="load_npy" class="doc_header"><code>load_npy</code><a href="https://githu
 
 
 <div class="output_markdown rendered_html output_subarea ">
-<h4 id="sample_file" class="doc_header"><code>sample_file</code><a href="https://github.com/mnpinto/dl_pipeline/tree/master/dl_pipeline/audio/core.py#L23" class="source_link" style="float:right">[source]</a></h4><blockquote><p><code>sample_file</code>()</p>
+<h4 id="sample_file" class="doc_header"><code>sample_file</code><a href="https://github.com/mnpinto/dl_pipeline/tree/master/dl_pipeline/audio/core.py#L26" class="source_link" style="float:right">[source]</a></h4><blockquote><p><code>sample_file</code>()</p>
 </blockquote>
 
 </div>
@@ -167,8 +167,8 @@ <h4 id="sample_file" class="doc_header"><code>sample_file</code><a href="https:/
 <div class="output_area">
 
 <div class="output_subarea output_stream output_stdout output_text">
-<pre>CPU times: user 15 ms, sys: 3.78 ms, total: 18.8 ms
-Wall time: 17.8 ms
+<pre>CPU times: user 19.5 ms, sys: 0 ns, total: 19.5 ms
+Wall time: 19.8 ms
 </pre>
 </div>
 </div>
@@ -178,7 +178,7 @@ <h4 id="sample_file" class="doc_header"><code>sample_file</code><a href="https:/
 
 
 <div class="output_text output_subarea output_execute_result">
-<pre>[&lt;matplotlib.lines.Line2D at 0x7f8ff32905e0&gt;]</pre>
+<pre>[&lt;matplotlib.lines.Line2D at 0x7f329ae146a0&gt;]</pre>
 </div>
 
 </div>
@@ -212,7 +212,7 @@ <h4 id="sample_file" class="doc_header"><code>sample_file</code><a href="https:/
 
 
 <div class="output_markdown rendered_html output_subarea ">
-<h4 id="melspectrogram" class="doc_header"><code>melspectrogram</code><a href="https://github.com/mnpinto/dl_pipeline/tree/master/dl_pipeline/audio/core.py#L27" class="source_link" style="float:right">[source]</a></h4><blockquote><p><code>melspectrogram</code>(<strong><code>wav</code></strong>, <strong><code>sample_rate</code></strong>=<em><code>22050</code></em>, <strong><code>n_fft</code></strong>=<em><code>2048</code></em>, <strong><code>n_mels</code></strong>=<em><code>128</code></em>, <strong><code>hop_length</code></strong>=<em><code>512</code></em>, <strong><code>window</code></strong>=<em><code>'hann'</code></em>, <strong><code>center</code></strong>=<em><code>True</code></em>, <strong><code>pad_mode</code></strong>=<em><code>'reflect'</code></em>, <strong><code>power</code></strong>=<em><code>2.0</code></em>, <strong><code>htk</code></strong>=<em><code>False</code></em>, <strong><code>fmin</code></strong>=<em><code>0.0</code></em>, <strong><code>fmax</code></strong>=<em><code>None</code></em>, <strong><code>norm</code></strong>=<em><code>1</code></em>, <strong><code>trainable_mel</code></strong>=<em><code>False</code></em>, <strong><code>trainable_STFT</code></strong>=<em><code>False</code></em>, <strong><code>verbose</code></strong>=<em><code>False</code></em>, <strong><code>eps</code></strong>=<em><code>1e-06</code></em>, <strong><code>cuda</code></strong>=<em><code>False</code></em>, <strong><code>log</code></strong>=<em><code>True</code></em>, <strong>**<code>kwargs</code></strong>)</p>
+<h4 id="melspectrogram" class="doc_header"><code>melspectrogram</code><a href="https://github.com/mnpinto/dl_pipeline/tree/master/dl_pipeline/audio/core.py#L30" class="source_link" style="float:right">[source]</a></h4><blockquote><p><code>melspectrogram</code>(<strong><code>wav</code></strong>, <strong><code>sample_rate</code></strong>=<em><code>22050</code></em>, <strong><code>n_fft</code></strong>=<em><code>2048</code></em>, <strong><code>n_mels</code></strong>=<em><code>128</code></em>, <strong><code>hop_length</code></strong>=<em><code>512</code></em>, <strong><code>window</code></strong>=<em><code>'hann'</code></em>, <strong><code>center</code></strong>=<em><code>True</code></em>, <strong><code>pad_mode</code></strong>=<em><code>'reflect'</code></em>, <strong><code>power</code></strong>=<em><code>2.0</code></em>, <strong><code>htk</code></strong>=<em><code>False</code></em>, <strong><code>fmin</code></strong>=<em><code>0.0</code></em>, <strong><code>fmax</code></strong>=<em><code>None</code></em>, <strong><code>norm</code></strong>=<em><code>1</code></em>, <strong><code>trainable_mel</code></strong>=<em><code>False</code></em>, <strong><code>trainable_STFT</code></strong>=<em><code>False</code></em>, <strong><code>verbose</code></strong>=<em><code>False</code></em>, <strong><code>eps</code></strong>=<em><code>1e-06</code></em>, <strong><code>cuda</code></strong>=<em><code>False</code></em>, <strong><code>log</code></strong>=<em><code>True</code></em>, <strong>**<code>kwargs</code></strong>)</p>
 </blockquote>
 
 </div>
@@ -236,7 +236,7 @@ <h4 id="melspectrogram" class="doc_header"><code>melspectrogram</code><a href="h
 
 
 <div class="output_markdown rendered_html output_subarea ">
-<h4 id="show_sample" class="doc_header"><code>show_sample</code><a href="https://github.com/mnpinto/dl_pipeline/tree/master/dl_pipeline/audio/core.py#L38" class="source_link" style="float:right">[source]</a></h4><blockquote><p><code>show_sample</code>(<strong><code>file</code></strong>=<em><code>'/kaggle/kaggle_rainforest_audio/data/npy32000/train/000316da7.npy'</code></em>, <strong><code>tfms</code></strong>=<em><code>&lt;lambda&gt;</code></em>, <strong><code>sample_rate</code></strong>=<em><code>48000</code></em>, <strong><code>hop_length</code></strong>=<em><code>512</code></em>, <strong><code>width</code></strong>=<em><code>512</code></em>, <strong><code>figsize</code></strong>=<em><code>(10, 4)</code></em>, <strong><code>vmin</code></strong>=<em><code>-10</code></em>, <strong><code>vmax</code></strong>=<em><code>2</code></em>)</p>
+<h4 id="show_sample" class="doc_header"><code>show_sample</code><a href="https://github.com/mnpinto/dl_pipeline/tree/master/dl_pipeline/audio/core.py#L41" class="source_link" style="float:right">[source]</a></h4><blockquote><p><code>show_sample</code>(<strong><code>file</code></strong>=<em><code>'/kaggle/kaggle_rainforest_audio/data/npy32000/train/000316da7.npy'</code></em>, <strong><code>tfms</code></strong>=<em><code>&lt;lambda&gt;</code></em>, <strong><code>sample_rate</code></strong>=<em><code>48000</code></em>, <strong><code>hop_length</code></strong>=<em><code>512</code></em>, <strong><code>width</code></strong>=<em><code>512</code></em>, <strong><code>figsize</code></strong>=<em><code>(10, 4)</code></em>, <strong><code>vmin</code></strong>=<em><code>-10</code></em>, <strong><code>vmax</code></strong>=<em><code>2</code></em>)</p>
 </blockquote>
 
 </div>
@@ -289,8 +289,8 @@ <h4 id="show_sample" class="doc_header"><code>show_sample</code><a href="https:/
 <div class="output_area">
 
 <div class="output_subarea output_stream output_stdout output_text">
-<pre>CPU times: user 571 ms, sys: 1.7 ms, total: 572 ms
-Wall time: 368 ms
+<pre>CPU times: user 642 ms, sys: 40.4 ms, total: 683 ms
+Wall time: 410 ms
 </pre>
 </div>
 </div>
@@ -301,6 +301,67 @@ <h4 id="show_sample" class="doc_header"><code>show_sample</code><a href="https:/
 </div>
     {% endraw %}
 
+<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="Saving-audio-data-in-npy-files">Saving audio data in npy files<a class="anchor-link" href="#Saving-audio-data-in-npy-files"> </a></h3>
+</div>
+</div>
+</div>
+    {% raw %}
+
+<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="audio2npy" class="doc_header"><code>audio2npy</code><a href="https://github.com/mnpinto/dl_pipeline/tree/master/dl_pipeline/audio/core.py#L50" class="source_link" style="float:right">[source]</a></h4><blockquote><p><code>audio2npy</code>(<strong><code>file</code></strong>, <strong><code>path_save</code></strong>, <strong><code>sample_rate</code></strong>=<em><code>32000</code></em>)</p>
+</blockquote>
+
+</div>
+
+</div>
+
+</div>
+</div>
+
+</div>
+    {% endraw %}
+
+    {% raw %}
+
+<div class="cell border-box-sizing code_cell rendered">
+
+</div>
+    {% endraw %}
+
+<div class="cell border-box-sizing text_cell rendered"><div class="inner_cell">
+<div class="text_cell_render border-box-sizing rendered_html">
+<div class="highlight"><pre><span></span><span class="n">sample_rate</span> <span class="o">=</span> <span class="mi">32_000</span>
+<span class="n">path</span> <span class="o">=</span> <span class="n">Path</span><span class="p">(</span><span class="s1">&#39;/kaggle/kaggle_rainforest_audio/data&#39;</span><span class="p">)</span>
+<span class="n">train_path</span> <span class="o">=</span> <span class="n">path</span><span class="o">/</span><span class="s1">&#39;train&#39;</span>
+<span class="n">test_path</span> <span class="o">=</span> <span class="n">path</span><span class="o">/</span><span class="s1">&#39;test&#39;</span>
+<span class="n">train_path_npy</span> <span class="o">=</span> <span class="n">Path</span><span class="p">(</span><span class="n">path</span><span class="o">/</span><span class="sa">f</span><span class="s1">&#39;npy</span><span class="si">{</span><span class="n">sample_rate</span><span class="si">}</span><span class="s1">/train&#39;</span><span class="p">)</span>
+<span class="n">test_path_npy</span> <span class="o">=</span> <span class="n">Path</span><span class="p">(</span><span class="n">path</span><span class="o">/</span><span class="sa">f</span><span class="s1">&#39;npy</span><span class="si">{</span><span class="n">sample_rate</span><span class="si">}</span><span class="s1">/test&#39;</span><span class="p">)</span>
+
+<span class="n">files</span> <span class="o">=</span> <span class="n">train_path</span><span class="o">.</span><span class="n">ls</span><span class="p">()</span>
+<span class="n">files</span> <span class="o">=</span> <span class="p">[</span><span class="n">f</span> <span class="k">for</span> <span class="n">f</span> <span class="ow">in</span> <span class="n">files</span> <span class="k">if</span> <span class="s1">&#39;.flac&#39;</span> <span class="ow">in</span> <span class="n">f</span><span class="o">.</span><span class="n">suffix</span><span class="p">]</span>
+
+<span class="n">f</span> <span class="o">=</span> <span class="n">partial</span><span class="p">(</span><span class="n">audio2npy</span><span class="p">,</span> <span class="n">path_save</span><span class="o">=</span><span class="n">train_path_npy</span><span class="p">,</span> <span class="n">sample_rate</span><span class="o">=</span><span class="n">sample_rate</span><span class="p">)</span>
+<span class="n">parallel</span><span class="p">(</span><span class="n">f</span><span class="p">,</span> <span class="n">files</span><span class="p">)</span>
+
+<span class="n">files</span> <span class="o">=</span> <span class="n">test_path</span><span class="o">.</span><span class="n">ls</span><span class="p">()</span>
+<span class="n">files</span> <span class="o">=</span> <span class="p">[</span><span class="n">f</span> <span class="k">for</span> <span class="n">f</span> <span class="ow">in</span> <span class="n">files</span> <span class="k">if</span> <span class="s1">&#39;.flac&#39;</span> <span class="ow">in</span> <span class="n">f</span><span class="o">.</span><span class="n">suffix</span><span class="p">]</span>
+<span class="n">f</span> <span class="o">=</span> <span class="n">partial</span><span class="p">(</span><span class="n">audio2npy</span><span class="p">,</span> <span class="n">path_save</span><span class="o">=</span><span class="n">test_path_npy</span><span class="p">,</span> <span class="n">sample_rate</span><span class="o">=</span><span class="n">sample_rate</span><span class="p">)</span>
+<span class="n">parallel</span><span class="p">(</span><span class="n">f</span><span class="p">,</span> <span class="n">files</span><span class="p">)</span>
+</pre></div>
+
+</div>
+</div>
+</div>
 </div>