facebookresearch--audiocraft
Alexandre Défossez (autodoc)
630 行
37 KiB
HTML
630 行
37 KiB
HTML
<!doctype html>
|
||
<html lang="en">
|
||
<head>
|
||
<meta charset="utf-8">
|
||
<meta name="viewport" content="width=device-width, initial-scale=1, minimum-scale=1">
|
||
<meta name="generator" content="pdoc3 0.11.5">
|
||
<title>audiocraft.modules.seanet API documentation</title>
|
||
<meta name="description" content="">
|
||
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/13.0.0/sanitize.min.css" integrity="sha512-y1dtMcuvtTMJc1yPgEqF0ZjQbhnc/bFhyvIyVNb9Zk5mIGtqVaAB1Ttl28su8AvFMOY0EwRbAe+HCLqj6W7/KA==" crossorigin>
|
||
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/10up-sanitize.css/13.0.0/typography.min.css" integrity="sha512-Y1DYSb995BAfxobCkKepB1BqJJTPrOp3zPL74AWFugHHmmdcvO+C48WLrUOlhGMc0QG7AE3f7gmvvcrmX2fDoA==" crossorigin>
|
||
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/11.9.0/styles/default.min.css" crossorigin>
|
||
<style>:root{--highlight-color:#fe9}.flex{display:flex !important}body{line-height:1.5em}#content{padding:20px}#sidebar{padding:1.5em;overflow:hidden}#sidebar > *:last-child{margin-bottom:2cm}.http-server-breadcrumbs{font-size:130%;margin:0 0 15px 0}#footer{font-size:.75em;padding:5px 30px;border-top:1px solid #ddd;text-align:right}#footer p{margin:0 0 0 1em;display:inline-block}#footer p:last-child{margin-right:30px}h1,h2,h3,h4,h5{font-weight:300}h1{font-size:2.5em;line-height:1.1em}h2{font-size:1.75em;margin:2em 0 .50em 0}h3{font-size:1.4em;margin:1.6em 0 .7em 0}h4{margin:0;font-size:105%}h1:target,h2:target,h3:target,h4:target,h5:target,h6:target{background:var(--highlight-color);padding:.2em 0}a{color:#058;text-decoration:none;transition:color .2s ease-in-out}a:visited{color:#503}a:hover{color:#b62}.title code{font-weight:bold}h2[id^="header-"]{margin-top:2em}.ident{color:#900;font-weight:bold}pre code{font-size:.8em;line-height:1.4em;padding:1em;display:block}code{background:#f3f3f3;font-family:"DejaVu Sans Mono",monospace;padding:1px 4px;overflow-wrap:break-word}h1 code{background:transparent}pre{border-top:1px solid #ccc;border-bottom:1px solid #ccc;margin:1em 0}#http-server-module-list{display:flex;flex-flow:column}#http-server-module-list div{display:flex}#http-server-module-list dt{min-width:10%}#http-server-module-list p{margin-top:0}.toc ul,#index{list-style-type:none;margin:0;padding:0}#index code{background:transparent}#index h3{border-bottom:1px solid #ddd}#index ul{padding:0}#index h4{margin-top:.6em;font-weight:bold}@media (min-width:200ex){#index .two-column{column-count:2}}@media (min-width:300ex){#index .two-column{column-count:3}}dl{margin-bottom:2em}dl dl:last-child{margin-bottom:4em}dd{margin:0 0 1em 3em}#header-classes + dl > dd{margin-bottom:3em}dd dd{margin-left:2em}dd p{margin:10px 0}.name{background:#eee;font-size:.85em;padding:5px 10px;display:inline-block;min-width:40%}.name:hover{background:#e0e0e0}dt:target .name{background:var(--highlight-color)}.name > span:first-child{white-space:nowrap}.name.class > span:nth-child(2){margin-left:.4em}.inherited{color:#999;border-left:5px solid #eee;padding-left:1em}.inheritance em{font-style:normal;font-weight:bold}.desc h2{font-weight:400;font-size:1.25em}.desc h3{font-size:1em}.desc dt code{background:inherit}.source > summary,.git-link-div{color:#666;text-align:right;font-weight:400;font-size:.8em;text-transform:uppercase}.source summary > *{white-space:nowrap;cursor:pointer}.git-link{color:inherit;margin-left:1em}.source pre{max-height:500px;overflow:auto;margin:0}.source pre code{font-size:12px;overflow:visible;min-width:max-content}.hlist{list-style:none}.hlist li{display:inline}.hlist li:after{content:',\2002'}.hlist li:last-child:after{content:none}.hlist .hlist{display:inline;padding-left:1em}img{max-width:100%}td{padding:0 .5em}.admonition{padding:.1em 1em;margin:1em 0}.admonition-title{font-weight:bold}.admonition.note,.admonition.info,.admonition.important{background:#aef}.admonition.todo,.admonition.versionadded,.admonition.tip,.admonition.hint{background:#dfd}.admonition.warning,.admonition.versionchanged,.admonition.deprecated{background:#fd4}.admonition.error,.admonition.danger,.admonition.caution{background:lightpink}</style>
|
||
<style media="screen and (min-width: 700px)">@media screen and (min-width:700px){#sidebar{width:30%;height:100vh;overflow:auto;position:sticky;top:0}#content{width:70%;max-width:100ch;padding:3em 4em;border-left:1px solid #ddd}pre code{font-size:1em}.name{font-size:1em}main{display:flex;flex-direction:row-reverse;justify-content:flex-end}.toc ul ul,#index ul ul{padding-left:1em}.toc > ul > li{margin-top:.5em}}</style>
|
||
<style media="print">@media print{#sidebar h1{page-break-before:always}.source{display:none}}@media print{*{background:transparent !important;color:#000 !important;box-shadow:none !important;text-shadow:none !important}a[href]:after{content:" (" attr(href) ")";font-size:90%}a[href][title]:after{content:none}abbr[title]:after{content:" (" attr(title) ")"}.ir a:after,a[href^="javascript:"]:after,a[href^="#"]:after{content:""}pre,blockquote{border:1px solid #999;page-break-inside:avoid}thead{display:table-header-group}tr,img{page-break-inside:avoid}img{max-width:100% !important}@page{margin:0.5cm}p,h2,h3{orphans:3;widows:3}h1,h2,h3,h4,h5,h6{page-break-after:avoid}}</style>
|
||
<script defer src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/11.9.0/highlight.min.js" integrity="sha512-D9gUyxqja7hBtkWpPWGt9wfbfaMGVt9gnyCvYa+jojwwPHLCzUm5i8rpk7vD7wNee9bA35eYIjobYPaQuKS1MQ==" crossorigin></script>
|
||
<script>window.addEventListener('DOMContentLoaded', () => {
|
||
hljs.configure({languages: ['bash', 'css', 'diff', 'graphql', 'ini', 'javascript', 'json', 'plaintext', 'python', 'python-repl', 'rust', 'shell', 'sql', 'typescript', 'xml', 'yaml']});
|
||
hljs.highlightAll();
|
||
/* Collapse source docstrings */
|
||
setTimeout(() => {
|
||
[...document.querySelectorAll('.hljs.language-python > .hljs-string')]
|
||
.filter(el => el.innerHTML.length > 200 && ['"""', "'''"].includes(el.innerHTML.substring(0, 3)))
|
||
.forEach(el => {
|
||
let d = document.createElement('details');
|
||
d.classList.add('hljs-string');
|
||
d.innerHTML = '<summary>"""</summary>' + el.innerHTML.substring(3);
|
||
el.replaceWith(d);
|
||
});
|
||
}, 100);
|
||
})</script>
|
||
</head>
|
||
<body>
|
||
<main>
|
||
<article id="content">
|
||
<header>
|
||
<h1 class="title">Module <code>audiocraft.modules.seanet</code></h1>
|
||
</header>
|
||
<section id="section-intro">
|
||
</section>
|
||
<section>
|
||
</section>
|
||
<section>
|
||
</section>
|
||
<section>
|
||
</section>
|
||
<section>
|
||
<h2 class="section-title" id="header-classes">Classes</h2>
|
||
<dl>
|
||
<dt id="audiocraft.modules.seanet.SEANetDecoder"><code class="flex name class">
|
||
<span>class <span class="ident">SEANetDecoder</span></span>
|
||
<span>(</span><span>channels: int = 1,<br>dimension: int = 128,<br>n_filters: int = 32,<br>n_residual_layers: int = 3,<br>ratios: List[int] = [8, 5, 4, 2],<br>activation: str = 'ELU',<br>activation_params: dict = {'alpha': 1.0},<br>final_activation: str | None = None,<br>final_activation_params: dict | None = None,<br>norm: str = 'none',<br>norm_params: Dict[str, Any] = {},<br>kernel_size: int = 7,<br>last_kernel_size: int = 7,<br>residual_kernel_size: int = 3,<br>dilation_base: int = 2,<br>causal: bool = False,<br>pad_mode: str = 'reflect',<br>true_skip: bool = True,<br>compress: int = 2,<br>lstm: int = 0,<br>disable_norm_outer_blocks: int = 0,<br>trim_right_ratio: float = 1.0)</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">class SEANetDecoder(nn.Module):
|
||
"""SEANet decoder.
|
||
|
||
Args:
|
||
channels (int): Audio channels.
|
||
dimension (int): Intermediate representation dimension.
|
||
n_filters (int): Base width for the model.
|
||
n_residual_layers (int): nb of residual layers.
|
||
ratios (Sequence[int]): kernel size and stride ratios.
|
||
activation (str): Activation function.
|
||
activation_params (dict): Parameters to provide to the activation function.
|
||
final_activation (str): Final activation function after all convolutions.
|
||
final_activation_params (dict): Parameters to provide to the activation function.
|
||
norm (str): Normalization method.
|
||
norm_params (dict): Parameters to provide to the underlying normalization used along with the convolution.
|
||
kernel_size (int): Kernel size for the initial convolution.
|
||
last_kernel_size (int): Kernel size for the initial convolution.
|
||
residual_kernel_size (int): Kernel size for the residual layers.
|
||
dilation_base (int): How much to increase the dilation with each layer.
|
||
causal (bool): Whether to use fully causal convolution.
|
||
pad_mode (str): Padding mode for the convolutions.
|
||
true_skip (bool): Whether to use true skip connection or a simple.
|
||
(streamable) convolution as the skip connection in the residual network blocks.
|
||
compress (int): Reduced dimensionality in residual branches (from Demucs v3).
|
||
lstm (int): Number of LSTM layers at the end of the encoder.
|
||
disable_norm_outer_blocks (int): Number of blocks for which we don't apply norm.
|
||
For the decoder, it corresponds to the N last blocks.
|
||
trim_right_ratio (float): Ratio for trimming at the right of the transposed convolution under the causal setup.
|
||
If equal to 1.0, it means that all the trimming is done at the right.
|
||
"""
|
||
def __init__(self, channels: int = 1, dimension: int = 128, n_filters: int = 32, n_residual_layers: int = 3,
|
||
ratios: tp.List[int] = [8, 5, 4, 2], activation: str = 'ELU', activation_params: dict = {'alpha': 1.0},
|
||
final_activation: tp.Optional[str] = None, final_activation_params: tp.Optional[dict] = None,
|
||
norm: str = 'none', norm_params: tp.Dict[str, tp.Any] = {}, kernel_size: int = 7,
|
||
last_kernel_size: int = 7, residual_kernel_size: int = 3, dilation_base: int = 2, causal: bool = False,
|
||
pad_mode: str = 'reflect', true_skip: bool = True, compress: int = 2, lstm: int = 0,
|
||
disable_norm_outer_blocks: int = 0, trim_right_ratio: float = 1.0):
|
||
super().__init__()
|
||
self.dimension = dimension
|
||
self.channels = channels
|
||
self.n_filters = n_filters
|
||
self.ratios = ratios
|
||
del ratios
|
||
self.n_residual_layers = n_residual_layers
|
||
self.hop_length = np.prod(self.ratios)
|
||
self.n_blocks = len(self.ratios) + 2 # first and last conv + residual blocks
|
||
self.disable_norm_outer_blocks = disable_norm_outer_blocks
|
||
assert self.disable_norm_outer_blocks >= 0 and self.disable_norm_outer_blocks <= self.n_blocks, \
|
||
"Number of blocks for which to disable norm is invalid." \
|
||
"It should be lower or equal to the actual number of blocks in the network and greater or equal to 0."
|
||
|
||
act = getattr(nn, activation)
|
||
mult = int(2 ** len(self.ratios))
|
||
model: tp.List[nn.Module] = [
|
||
StreamableConv1d(dimension, mult * n_filters, kernel_size,
|
||
norm='none' if self.disable_norm_outer_blocks == self.n_blocks else norm,
|
||
norm_kwargs=norm_params, causal=causal, pad_mode=pad_mode)
|
||
]
|
||
|
||
if lstm:
|
||
model += [StreamableLSTM(mult * n_filters, num_layers=lstm)]
|
||
|
||
# Upsample to raw audio scale
|
||
for i, ratio in enumerate(self.ratios):
|
||
block_norm = 'none' if self.disable_norm_outer_blocks >= self.n_blocks - (i + 1) else norm
|
||
# Add upsampling layers
|
||
model += [
|
||
act(**activation_params),
|
||
StreamableConvTranspose1d(mult * n_filters, mult * n_filters // 2,
|
||
kernel_size=ratio * 2, stride=ratio,
|
||
norm=block_norm, norm_kwargs=norm_params,
|
||
causal=causal, trim_right_ratio=trim_right_ratio),
|
||
]
|
||
# Add residual layers
|
||
for j in range(n_residual_layers):
|
||
model += [
|
||
SEANetResnetBlock(mult * n_filters // 2, kernel_sizes=[residual_kernel_size, 1],
|
||
dilations=[dilation_base ** j, 1],
|
||
activation=activation, activation_params=activation_params,
|
||
norm=block_norm, norm_params=norm_params, causal=causal,
|
||
pad_mode=pad_mode, compress=compress, true_skip=true_skip)]
|
||
|
||
mult //= 2
|
||
|
||
# Add final layers
|
||
model += [
|
||
act(**activation_params),
|
||
StreamableConv1d(n_filters, channels, last_kernel_size,
|
||
norm='none' if self.disable_norm_outer_blocks >= 1 else norm,
|
||
norm_kwargs=norm_params, causal=causal, pad_mode=pad_mode)
|
||
]
|
||
# Add optional final activation to decoder (eg. tanh)
|
||
if final_activation is not None:
|
||
final_act = getattr(nn, final_activation)
|
||
final_activation_params = final_activation_params or {}
|
||
model += [
|
||
final_act(**final_activation_params)
|
||
]
|
||
self.model = nn.Sequential(*model)
|
||
|
||
def forward(self, z):
|
||
y = self.model(z)
|
||
return y</code></pre>
|
||
</details>
|
||
<div class="desc"><p>SEANet decoder.</p>
|
||
<h2 id="args">Args</h2>
|
||
<dl>
|
||
<dt><strong><code>channels</code></strong> : <code>int</code></dt>
|
||
<dd>Audio channels.</dd>
|
||
<dt><strong><code>dimension</code></strong> : <code>int</code></dt>
|
||
<dd>Intermediate representation dimension.</dd>
|
||
<dt><strong><code>n_filters</code></strong> : <code>int</code></dt>
|
||
<dd>Base width for the model.</dd>
|
||
<dt><strong><code>n_residual_layers</code></strong> : <code>int</code></dt>
|
||
<dd>nb of residual layers.</dd>
|
||
<dt><strong><code>ratios</code></strong> : <code>Sequence[int]</code></dt>
|
||
<dd>kernel size and stride ratios.</dd>
|
||
<dt><strong><code>activation</code></strong> : <code>str</code></dt>
|
||
<dd>Activation function.</dd>
|
||
<dt><strong><code>activation_params</code></strong> : <code>dict</code></dt>
|
||
<dd>Parameters to provide to the activation function.</dd>
|
||
<dt><strong><code>final_activation</code></strong> : <code>str</code></dt>
|
||
<dd>Final activation function after all convolutions.</dd>
|
||
<dt><strong><code>final_activation_params</code></strong> : <code>dict</code></dt>
|
||
<dd>Parameters to provide to the activation function.</dd>
|
||
<dt><strong><code>norm</code></strong> : <code>str</code></dt>
|
||
<dd>Normalization method.</dd>
|
||
<dt><strong><code>norm_params</code></strong> : <code>dict</code></dt>
|
||
<dd>Parameters to provide to the underlying normalization used along with the convolution.</dd>
|
||
<dt><strong><code>kernel_size</code></strong> : <code>int</code></dt>
|
||
<dd>Kernel size for the initial convolution.</dd>
|
||
<dt><strong><code>last_kernel_size</code></strong> : <code>int</code></dt>
|
||
<dd>Kernel size for the initial convolution.</dd>
|
||
<dt><strong><code>residual_kernel_size</code></strong> : <code>int</code></dt>
|
||
<dd>Kernel size for the residual layers.</dd>
|
||
<dt><strong><code>dilation_base</code></strong> : <code>int</code></dt>
|
||
<dd>How much to increase the dilation with each layer.</dd>
|
||
<dt><strong><code>causal</code></strong> : <code>bool</code></dt>
|
||
<dd>Whether to use fully causal convolution.</dd>
|
||
<dt><strong><code>pad_mode</code></strong> : <code>str</code></dt>
|
||
<dd>Padding mode for the convolutions.</dd>
|
||
<dt><strong><code>true_skip</code></strong> : <code>bool</code></dt>
|
||
<dd>Whether to use true skip connection or a simple.
|
||
(streamable) convolution as the skip connection in the residual network blocks.</dd>
|
||
<dt><strong><code>compress</code></strong> : <code>int</code></dt>
|
||
<dd>Reduced dimensionality in residual branches (from Demucs v3).</dd>
|
||
<dt><strong><code>lstm</code></strong> : <code>int</code></dt>
|
||
<dd>Number of LSTM layers at the end of the encoder.</dd>
|
||
<dt><strong><code>disable_norm_outer_blocks</code></strong> : <code>int</code></dt>
|
||
<dd>Number of blocks for which we don't apply norm.
|
||
For the decoder, it corresponds to the N last blocks.</dd>
|
||
<dt><strong><code>trim_right_ratio</code></strong> : <code>float</code></dt>
|
||
<dd>Ratio for trimming at the right of the transposed convolution under the causal setup.
|
||
If equal to 1.0, it means that all the trimming is done at the right.</dd>
|
||
</dl>
|
||
<p>Initializes internal Module state, shared by both nn.Module and ScriptModule.</p></div>
|
||
<h3>Ancestors</h3>
|
||
<ul class="hlist">
|
||
<li>torch.nn.modules.module.Module</li>
|
||
</ul>
|
||
<h3>Class variables</h3>
|
||
<dl>
|
||
<dt id="audiocraft.modules.seanet.SEANetDecoder.call_super_init"><code class="name">var <span class="ident">call_super_init</span> : bool</code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
<dt id="audiocraft.modules.seanet.SEANetDecoder.dump_patches"><code class="name">var <span class="ident">dump_patches</span> : bool</code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
<dt id="audiocraft.modules.seanet.SEANetDecoder.training"><code class="name">var <span class="ident">training</span> : bool</code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
</dl>
|
||
<h3>Methods</h3>
|
||
<dl>
|
||
<dt id="audiocraft.modules.seanet.SEANetDecoder.forward"><code class="name flex">
|
||
<span>def <span class="ident">forward</span></span>(<span>self, z) ‑> Callable[..., Any]</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def forward(self, z):
|
||
y = self.model(z)
|
||
return y</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Defines the computation performed at every call.</p>
|
||
<p>Should be overridden by all subclasses.</p>
|
||
<div class="admonition note">
|
||
<p class="admonition-title">Note</p>
|
||
<p>Although the recipe for forward pass needs to be defined within
|
||
this function, one should call the :class:<code>Module</code> instance afterwards
|
||
instead of this since the former takes care of running the
|
||
registered hooks while the latter silently ignores them.</p>
|
||
</div></div>
|
||
</dd>
|
||
</dl>
|
||
</dd>
|
||
<dt id="audiocraft.modules.seanet.SEANetEncoder"><code class="flex name class">
|
||
<span>class <span class="ident">SEANetEncoder</span></span>
|
||
<span>(</span><span>channels: int = 1,<br>dimension: int = 128,<br>n_filters: int = 32,<br>n_residual_layers: int = 3,<br>ratios: List[int] = [8, 5, 4, 2],<br>activation: str = 'ELU',<br>activation_params: dict = {'alpha': 1.0},<br>norm: str = 'none',<br>norm_params: Dict[str, Any] = {},<br>kernel_size: int = 7,<br>last_kernel_size: int = 7,<br>residual_kernel_size: int = 3,<br>dilation_base: int = 2,<br>causal: bool = False,<br>pad_mode: str = 'reflect',<br>true_skip: bool = True,<br>compress: int = 2,<br>lstm: int = 0,<br>disable_norm_outer_blocks: int = 0)</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">class SEANetEncoder(nn.Module):
|
||
"""SEANet encoder.
|
||
|
||
Args:
|
||
channels (int): Audio channels.
|
||
dimension (int): Intermediate representation dimension.
|
||
n_filters (int): Base width for the model.
|
||
n_residual_layers (int): nb of residual layers.
|
||
ratios (Sequence[int]): kernel size and stride ratios. The encoder uses downsampling ratios instead of
|
||
upsampling ratios, hence it will use the ratios in the reverse order to the ones specified here
|
||
that must match the decoder order. We use the decoder order as some models may only employ the decoder.
|
||
activation (str): Activation function.
|
||
activation_params (dict): Parameters to provide to the activation function.
|
||
norm (str): Normalization method.
|
||
norm_params (dict): Parameters to provide to the underlying normalization used along with the convolution.
|
||
kernel_size (int): Kernel size for the initial convolution.
|
||
last_kernel_size (int): Kernel size for the initial convolution.
|
||
residual_kernel_size (int): Kernel size for the residual layers.
|
||
dilation_base (int): How much to increase the dilation with each layer.
|
||
causal (bool): Whether to use fully causal convolution.
|
||
pad_mode (str): Padding mode for the convolutions.
|
||
true_skip (bool): Whether to use true skip connection or a simple
|
||
(streamable) convolution as the skip connection in the residual network blocks.
|
||
compress (int): Reduced dimensionality in residual branches (from Demucs v3).
|
||
lstm (int): Number of LSTM layers at the end of the encoder.
|
||
disable_norm_outer_blocks (int): Number of blocks for which we don't apply norm.
|
||
For the encoder, it corresponds to the N first blocks.
|
||
"""
|
||
def __init__(self, channels: int = 1, dimension: int = 128, n_filters: int = 32, n_residual_layers: int = 3,
|
||
ratios: tp.List[int] = [8, 5, 4, 2], activation: str = 'ELU', activation_params: dict = {'alpha': 1.0},
|
||
norm: str = 'none', norm_params: tp.Dict[str, tp.Any] = {}, kernel_size: int = 7,
|
||
last_kernel_size: int = 7, residual_kernel_size: int = 3, dilation_base: int = 2, causal: bool = False,
|
||
pad_mode: str = 'reflect', true_skip: bool = True, compress: int = 2, lstm: int = 0,
|
||
disable_norm_outer_blocks: int = 0):
|
||
super().__init__()
|
||
self.channels = channels
|
||
self.dimension = dimension
|
||
self.n_filters = n_filters
|
||
self.ratios = list(reversed(ratios))
|
||
del ratios
|
||
self.n_residual_layers = n_residual_layers
|
||
self.hop_length = np.prod(self.ratios)
|
||
self.n_blocks = len(self.ratios) + 2 # first and last conv + residual blocks
|
||
self.disable_norm_outer_blocks = disable_norm_outer_blocks
|
||
assert self.disable_norm_outer_blocks >= 0 and self.disable_norm_outer_blocks <= self.n_blocks, \
|
||
"Number of blocks for which to disable norm is invalid." \
|
||
"It should be lower or equal to the actual number of blocks in the network and greater or equal to 0."
|
||
|
||
act = getattr(nn, activation)
|
||
mult = 1
|
||
model: tp.List[nn.Module] = [
|
||
StreamableConv1d(channels, mult * n_filters, kernel_size,
|
||
norm='none' if self.disable_norm_outer_blocks >= 1 else norm,
|
||
norm_kwargs=norm_params, causal=causal, pad_mode=pad_mode)
|
||
]
|
||
# Downsample to raw audio scale
|
||
for i, ratio in enumerate(self.ratios):
|
||
block_norm = 'none' if self.disable_norm_outer_blocks >= i + 2 else norm
|
||
# Add residual layers
|
||
for j in range(n_residual_layers):
|
||
model += [
|
||
SEANetResnetBlock(mult * n_filters, kernel_sizes=[residual_kernel_size, 1],
|
||
dilations=[dilation_base ** j, 1],
|
||
norm=block_norm, norm_params=norm_params,
|
||
activation=activation, activation_params=activation_params,
|
||
causal=causal, pad_mode=pad_mode, compress=compress, true_skip=true_skip)]
|
||
|
||
# Add downsampling layers
|
||
model += [
|
||
act(**activation_params),
|
||
StreamableConv1d(mult * n_filters, mult * n_filters * 2,
|
||
kernel_size=ratio * 2, stride=ratio,
|
||
norm=block_norm, norm_kwargs=norm_params,
|
||
causal=causal, pad_mode=pad_mode),
|
||
]
|
||
mult *= 2
|
||
|
||
if lstm:
|
||
model += [StreamableLSTM(mult * n_filters, num_layers=lstm)]
|
||
|
||
model += [
|
||
act(**activation_params),
|
||
StreamableConv1d(mult * n_filters, dimension, last_kernel_size,
|
||
norm='none' if self.disable_norm_outer_blocks == self.n_blocks else norm,
|
||
norm_kwargs=norm_params, causal=causal, pad_mode=pad_mode)
|
||
]
|
||
|
||
self.model = nn.Sequential(*model)
|
||
|
||
def forward(self, x):
|
||
return self.model(x)</code></pre>
|
||
</details>
|
||
<div class="desc"><p>SEANet encoder.</p>
|
||
<h2 id="args">Args</h2>
|
||
<dl>
|
||
<dt><strong><code>channels</code></strong> : <code>int</code></dt>
|
||
<dd>Audio channels.</dd>
|
||
<dt><strong><code>dimension</code></strong> : <code>int</code></dt>
|
||
<dd>Intermediate representation dimension.</dd>
|
||
<dt><strong><code>n_filters</code></strong> : <code>int</code></dt>
|
||
<dd>Base width for the model.</dd>
|
||
<dt><strong><code>n_residual_layers</code></strong> : <code>int</code></dt>
|
||
<dd>nb of residual layers.</dd>
|
||
<dt><strong><code>ratios</code></strong> : <code>Sequence[int]</code></dt>
|
||
<dd>kernel size and stride ratios. The encoder uses downsampling ratios instead of
|
||
upsampling ratios, hence it will use the ratios in the reverse order to the ones specified here
|
||
that must match the decoder order. We use the decoder order as some models may only employ the decoder.</dd>
|
||
<dt><strong><code>activation</code></strong> : <code>str</code></dt>
|
||
<dd>Activation function.</dd>
|
||
<dt><strong><code>activation_params</code></strong> : <code>dict</code></dt>
|
||
<dd>Parameters to provide to the activation function.</dd>
|
||
<dt><strong><code>norm</code></strong> : <code>str</code></dt>
|
||
<dd>Normalization method.</dd>
|
||
<dt><strong><code>norm_params</code></strong> : <code>dict</code></dt>
|
||
<dd>Parameters to provide to the underlying normalization used along with the convolution.</dd>
|
||
<dt><strong><code>kernel_size</code></strong> : <code>int</code></dt>
|
||
<dd>Kernel size for the initial convolution.</dd>
|
||
<dt><strong><code>last_kernel_size</code></strong> : <code>int</code></dt>
|
||
<dd>Kernel size for the initial convolution.</dd>
|
||
<dt><strong><code>residual_kernel_size</code></strong> : <code>int</code></dt>
|
||
<dd>Kernel size for the residual layers.</dd>
|
||
<dt><strong><code>dilation_base</code></strong> : <code>int</code></dt>
|
||
<dd>How much to increase the dilation with each layer.</dd>
|
||
<dt><strong><code>causal</code></strong> : <code>bool</code></dt>
|
||
<dd>Whether to use fully causal convolution.</dd>
|
||
<dt><strong><code>pad_mode</code></strong> : <code>str</code></dt>
|
||
<dd>Padding mode for the convolutions.</dd>
|
||
<dt><strong><code>true_skip</code></strong> : <code>bool</code></dt>
|
||
<dd>Whether to use true skip connection or a simple
|
||
(streamable) convolution as the skip connection in the residual network blocks.</dd>
|
||
<dt><strong><code>compress</code></strong> : <code>int</code></dt>
|
||
<dd>Reduced dimensionality in residual branches (from Demucs v3).</dd>
|
||
<dt><strong><code>lstm</code></strong> : <code>int</code></dt>
|
||
<dd>Number of LSTM layers at the end of the encoder.</dd>
|
||
<dt><strong><code>disable_norm_outer_blocks</code></strong> : <code>int</code></dt>
|
||
<dd>Number of blocks for which we don't apply norm.
|
||
For the encoder, it corresponds to the N first blocks.</dd>
|
||
</dl>
|
||
<p>Initializes internal Module state, shared by both nn.Module and ScriptModule.</p></div>
|
||
<h3>Ancestors</h3>
|
||
<ul class="hlist">
|
||
<li>torch.nn.modules.module.Module</li>
|
||
</ul>
|
||
<h3>Class variables</h3>
|
||
<dl>
|
||
<dt id="audiocraft.modules.seanet.SEANetEncoder.call_super_init"><code class="name">var <span class="ident">call_super_init</span> : bool</code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
<dt id="audiocraft.modules.seanet.SEANetEncoder.dump_patches"><code class="name">var <span class="ident">dump_patches</span> : bool</code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
<dt id="audiocraft.modules.seanet.SEANetEncoder.training"><code class="name">var <span class="ident">training</span> : bool</code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
</dl>
|
||
<h3>Methods</h3>
|
||
<dl>
|
||
<dt id="audiocraft.modules.seanet.SEANetEncoder.forward"><code class="name flex">
|
||
<span>def <span class="ident">forward</span></span>(<span>self, x) ‑> Callable[..., Any]</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def forward(self, x):
|
||
return self.model(x)</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Defines the computation performed at every call.</p>
|
||
<p>Should be overridden by all subclasses.</p>
|
||
<div class="admonition note">
|
||
<p class="admonition-title">Note</p>
|
||
<p>Although the recipe for forward pass needs to be defined within
|
||
this function, one should call the :class:<code>Module</code> instance afterwards
|
||
instead of this since the former takes care of running the
|
||
registered hooks while the latter silently ignores them.</p>
|
||
</div></div>
|
||
</dd>
|
||
</dl>
|
||
</dd>
|
||
<dt id="audiocraft.modules.seanet.SEANetResnetBlock"><code class="flex name class">
|
||
<span>class <span class="ident">SEANetResnetBlock</span></span>
|
||
<span>(</span><span>dim: int,<br>kernel_sizes: List[int] = [3, 1],<br>dilations: List[int] = [1, 1],<br>activation: str = 'ELU',<br>activation_params: dict = {'alpha': 1.0},<br>norm: str = 'none',<br>norm_params: Dict[str, Any] = {},<br>causal: bool = False,<br>pad_mode: str = 'reflect',<br>compress: int = 2,<br>true_skip: bool = True)</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">class SEANetResnetBlock(nn.Module):
|
||
"""Residual block from SEANet model.
|
||
|
||
Args:
|
||
dim (int): Dimension of the input/output.
|
||
kernel_sizes (list): List of kernel sizes for the convolutions.
|
||
dilations (list): List of dilations for the convolutions.
|
||
activation (str): Activation function.
|
||
activation_params (dict): Parameters to provide to the activation function.
|
||
norm (str): Normalization method.
|
||
norm_params (dict): Parameters to provide to the underlying normalization used along with the convolution.
|
||
causal (bool): Whether to use fully causal convolution.
|
||
pad_mode (str): Padding mode for the convolutions.
|
||
compress (int): Reduced dimensionality in residual branches (from Demucs v3).
|
||
true_skip (bool): Whether to use true skip connection or a simple
|
||
(streamable) convolution as the skip connection.
|
||
"""
|
||
def __init__(self, dim: int, kernel_sizes: tp.List[int] = [3, 1], dilations: tp.List[int] = [1, 1],
|
||
activation: str = 'ELU', activation_params: dict = {'alpha': 1.0},
|
||
norm: str = 'none', norm_params: tp.Dict[str, tp.Any] = {}, causal: bool = False,
|
||
pad_mode: str = 'reflect', compress: int = 2, true_skip: bool = True):
|
||
super().__init__()
|
||
assert len(kernel_sizes) == len(dilations), 'Number of kernel sizes should match number of dilations'
|
||
act = getattr(nn, activation)
|
||
hidden = dim // compress
|
||
block = []
|
||
for i, (kernel_size, dilation) in enumerate(zip(kernel_sizes, dilations)):
|
||
in_chs = dim if i == 0 else hidden
|
||
out_chs = dim if i == len(kernel_sizes) - 1 else hidden
|
||
block += [
|
||
act(**activation_params),
|
||
StreamableConv1d(in_chs, out_chs, kernel_size=kernel_size, dilation=dilation,
|
||
norm=norm, norm_kwargs=norm_params,
|
||
causal=causal, pad_mode=pad_mode),
|
||
]
|
||
self.block = nn.Sequential(*block)
|
||
self.shortcut: nn.Module
|
||
if true_skip:
|
||
self.shortcut = nn.Identity()
|
||
else:
|
||
self.shortcut = StreamableConv1d(dim, dim, kernel_size=1, norm=norm, norm_kwargs=norm_params,
|
||
causal=causal, pad_mode=pad_mode)
|
||
|
||
def forward(self, x):
|
||
return self.shortcut(x) + self.block(x)</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Residual block from SEANet model.</p>
|
||
<h2 id="args">Args</h2>
|
||
<dl>
|
||
<dt><strong><code>dim</code></strong> : <code>int</code></dt>
|
||
<dd>Dimension of the input/output.</dd>
|
||
<dt><strong><code>kernel_sizes</code></strong> : <code>list</code></dt>
|
||
<dd>List of kernel sizes for the convolutions.</dd>
|
||
<dt><strong><code>dilations</code></strong> : <code>list</code></dt>
|
||
<dd>List of dilations for the convolutions.</dd>
|
||
<dt><strong><code>activation</code></strong> : <code>str</code></dt>
|
||
<dd>Activation function.</dd>
|
||
<dt><strong><code>activation_params</code></strong> : <code>dict</code></dt>
|
||
<dd>Parameters to provide to the activation function.</dd>
|
||
<dt><strong><code>norm</code></strong> : <code>str</code></dt>
|
||
<dd>Normalization method.</dd>
|
||
<dt><strong><code>norm_params</code></strong> : <code>dict</code></dt>
|
||
<dd>Parameters to provide to the underlying normalization used along with the convolution.</dd>
|
||
<dt><strong><code>causal</code></strong> : <code>bool</code></dt>
|
||
<dd>Whether to use fully causal convolution.</dd>
|
||
<dt><strong><code>pad_mode</code></strong> : <code>str</code></dt>
|
||
<dd>Padding mode for the convolutions.</dd>
|
||
<dt><strong><code>compress</code></strong> : <code>int</code></dt>
|
||
<dd>Reduced dimensionality in residual branches (from Demucs v3).</dd>
|
||
<dt><strong><code>true_skip</code></strong> : <code>bool</code></dt>
|
||
<dd>Whether to use true skip connection or a simple
|
||
(streamable) convolution as the skip connection.</dd>
|
||
</dl>
|
||
<p>Initializes internal Module state, shared by both nn.Module and ScriptModule.</p></div>
|
||
<h3>Ancestors</h3>
|
||
<ul class="hlist">
|
||
<li>torch.nn.modules.module.Module</li>
|
||
</ul>
|
||
<h3>Class variables</h3>
|
||
<dl>
|
||
<dt id="audiocraft.modules.seanet.SEANetResnetBlock.call_super_init"><code class="name">var <span class="ident">call_super_init</span> : bool</code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
<dt id="audiocraft.modules.seanet.SEANetResnetBlock.dump_patches"><code class="name">var <span class="ident">dump_patches</span> : bool</code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
<dt id="audiocraft.modules.seanet.SEANetResnetBlock.training"><code class="name">var <span class="ident">training</span> : bool</code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
</dl>
|
||
<h3>Methods</h3>
|
||
<dl>
|
||
<dt id="audiocraft.modules.seanet.SEANetResnetBlock.forward"><code class="name flex">
|
||
<span>def <span class="ident">forward</span></span>(<span>self, x) ‑> Callable[..., Any]</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def forward(self, x):
|
||
return self.shortcut(x) + self.block(x)</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Defines the computation performed at every call.</p>
|
||
<p>Should be overridden by all subclasses.</p>
|
||
<div class="admonition note">
|
||
<p class="admonition-title">Note</p>
|
||
<p>Although the recipe for forward pass needs to be defined within
|
||
this function, one should call the :class:<code>Module</code> instance afterwards
|
||
instead of this since the former takes care of running the
|
||
registered hooks while the latter silently ignores them.</p>
|
||
</div></div>
|
||
</dd>
|
||
</dl>
|
||
</dd>
|
||
</dl>
|
||
</section>
|
||
</article>
|
||
<nav id="sidebar">
|
||
<div class="toc">
|
||
<ul></ul>
|
||
</div>
|
||
<ul id="index">
|
||
<li><h3>Super-module</h3>
|
||
<ul>
|
||
<li><code><a title="audiocraft.modules" href="index.html">audiocraft.modules</a></code></li>
|
||
</ul>
|
||
</li>
|
||
<li><h3><a href="#header-classes">Classes</a></h3>
|
||
<ul>
|
||
<li>
|
||
<h4><code><a title="audiocraft.modules.seanet.SEANetDecoder" href="#audiocraft.modules.seanet.SEANetDecoder">SEANetDecoder</a></code></h4>
|
||
<ul class="">
|
||
<li><code><a title="audiocraft.modules.seanet.SEANetDecoder.call_super_init" href="#audiocraft.modules.seanet.SEANetDecoder.call_super_init">call_super_init</a></code></li>
|
||
<li><code><a title="audiocraft.modules.seanet.SEANetDecoder.dump_patches" href="#audiocraft.modules.seanet.SEANetDecoder.dump_patches">dump_patches</a></code></li>
|
||
<li><code><a title="audiocraft.modules.seanet.SEANetDecoder.forward" href="#audiocraft.modules.seanet.SEANetDecoder.forward">forward</a></code></li>
|
||
<li><code><a title="audiocraft.modules.seanet.SEANetDecoder.training" href="#audiocraft.modules.seanet.SEANetDecoder.training">training</a></code></li>
|
||
</ul>
|
||
</li>
|
||
<li>
|
||
<h4><code><a title="audiocraft.modules.seanet.SEANetEncoder" href="#audiocraft.modules.seanet.SEANetEncoder">SEANetEncoder</a></code></h4>
|
||
<ul class="">
|
||
<li><code><a title="audiocraft.modules.seanet.SEANetEncoder.call_super_init" href="#audiocraft.modules.seanet.SEANetEncoder.call_super_init">call_super_init</a></code></li>
|
||
<li><code><a title="audiocraft.modules.seanet.SEANetEncoder.dump_patches" href="#audiocraft.modules.seanet.SEANetEncoder.dump_patches">dump_patches</a></code></li>
|
||
<li><code><a title="audiocraft.modules.seanet.SEANetEncoder.forward" href="#audiocraft.modules.seanet.SEANetEncoder.forward">forward</a></code></li>
|
||
<li><code><a title="audiocraft.modules.seanet.SEANetEncoder.training" href="#audiocraft.modules.seanet.SEANetEncoder.training">training</a></code></li>
|
||
</ul>
|
||
</li>
|
||
<li>
|
||
<h4><code><a title="audiocraft.modules.seanet.SEANetResnetBlock" href="#audiocraft.modules.seanet.SEANetResnetBlock">SEANetResnetBlock</a></code></h4>
|
||
<ul class="">
|
||
<li><code><a title="audiocraft.modules.seanet.SEANetResnetBlock.call_super_init" href="#audiocraft.modules.seanet.SEANetResnetBlock.call_super_init">call_super_init</a></code></li>
|
||
<li><code><a title="audiocraft.modules.seanet.SEANetResnetBlock.dump_patches" href="#audiocraft.modules.seanet.SEANetResnetBlock.dump_patches">dump_patches</a></code></li>
|
||
<li><code><a title="audiocraft.modules.seanet.SEANetResnetBlock.forward" href="#audiocraft.modules.seanet.SEANetResnetBlock.forward">forward</a></code></li>
|
||
<li><code><a title="audiocraft.modules.seanet.SEANetResnetBlock.training" href="#audiocraft.modules.seanet.SEANetResnetBlock.training">training</a></code></li>
|
||
</ul>
|
||
</li>
|
||
</ul>
|
||
</li>
|
||
</ul>
|
||
</nav>
|
||
</main>
|
||
<footer id="footer">
|
||
<p>Generated by <a href="https://pdoc3.github.io/pdoc" title="pdoc: Python API documentation generator"><cite>pdoc</cite> 0.11.5</a>.</p>
|
||
</footer>
|
||
</body>
|
||
</html>
|