facebookresearch--audiocraft
Alexandre Défossez (autodoc)
800 行
37 KiB
HTML
800 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.models.unet API documentation</title>
|
||
<meta name="description" content="Pytorch Unet Module used for diffusion.">
|
||
<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.models.unet</code></h1>
|
||
</header>
|
||
<section id="section-intro">
|
||
<p>Pytorch Unet Module used for diffusion.</p>
|
||
</section>
|
||
<section>
|
||
</section>
|
||
<section>
|
||
</section>
|
||
<section>
|
||
<h2 class="section-title" id="header-functions">Functions</h2>
|
||
<dl>
|
||
<dt id="audiocraft.models.unet.get_model"><code class="name flex">
|
||
<span>def <span class="ident">get_model</span></span>(<span>cfg, channels: int, side: int, num_steps: int)</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_model(cfg, channels: int, side: int, num_steps: int):
|
||
if cfg.model == 'unet':
|
||
return DiffusionUnet(
|
||
chin=channels, num_steps=num_steps, **cfg.diffusion_unet)
|
||
else:
|
||
raise RuntimeError('Not Implemented')</code></pre>
|
||
</details>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
</dl>
|
||
</section>
|
||
<section>
|
||
<h2 class="section-title" id="header-classes">Classes</h2>
|
||
<dl>
|
||
<dt id="audiocraft.models.unet.BLSTM"><code class="flex name class">
|
||
<span>class <span class="ident">BLSTM</span></span>
|
||
<span>(</span><span>dim, layers=2)</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">class BLSTM(nn.Module):
|
||
"""BiLSTM with same hidden units as input dim.
|
||
"""
|
||
def __init__(self, dim, layers=2):
|
||
super().__init__()
|
||
self.lstm = nn.LSTM(bidirectional=True, num_layers=layers, hidden_size=dim, input_size=dim)
|
||
self.linear = nn.Linear(2 * dim, dim)
|
||
|
||
def forward(self, x):
|
||
x = x.permute(2, 0, 1)
|
||
x = self.lstm(x)[0]
|
||
x = self.linear(x)
|
||
x = x.permute(1, 2, 0)
|
||
return x</code></pre>
|
||
</details>
|
||
<div class="desc"><p>BiLSTM with same hidden units as input dim.</p>
|
||
<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.models.unet.BLSTM.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.models.unet.BLSTM.dump_patches"><code class="name">var <span class="ident">dump_patches</span> : bool</code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
<dt id="audiocraft.models.unet.BLSTM.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.models.unet.BLSTM.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):
|
||
x = x.permute(2, 0, 1)
|
||
x = self.lstm(x)[0]
|
||
x = self.linear(x)
|
||
x = x.permute(1, 2, 0)
|
||
return 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.models.unet.DecoderLayer"><code class="flex name class">
|
||
<span>class <span class="ident">DecoderLayer</span></span>
|
||
<span>(</span><span>chin: int,<br>chout: int,<br>kernel: int = 4,<br>stride: int = 2,<br>norm_groups: int = 4,<br>res_blocks: int = 1,<br>activation: Type[torch.nn.modules.module.Module] = torch.nn.modules.activation.ReLU,<br>dropout: float = 0.0)</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">class DecoderLayer(nn.Module):
|
||
def __init__(self, chin: int, chout: int, kernel: int = 4, stride: int = 2,
|
||
norm_groups: int = 4, res_blocks: int = 1, activation: tp.Type[nn.Module] = nn.ReLU,
|
||
dropout: float = 0.):
|
||
super().__init__()
|
||
padding = (kernel - stride) // 2
|
||
self.res_blocks = nn.Sequential(
|
||
*[ResBlock(chin, norm_groups=norm_groups, dilation=2**idx, dropout=dropout)
|
||
for idx in range(res_blocks)])
|
||
self.norm = nn.GroupNorm(norm_groups, chin)
|
||
ConvTr = nn.ConvTranspose1d
|
||
self.convtr = ConvTr(chin, chout, kernel, stride, padding, bias=False)
|
||
self.activation = activation()
|
||
|
||
def forward(self, x: torch.Tensor) -> torch.Tensor:
|
||
x = self.res_blocks(x)
|
||
x = self.norm(x)
|
||
x = self.activation(x)
|
||
x = self.convtr(x)
|
||
return x</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Base class for all neural network modules.</p>
|
||
<p>Your models should also subclass this class.</p>
|
||
<p>Modules can also contain other Modules, allowing to nest them in
|
||
a tree structure. You can assign the submodules as regular attributes::</p>
|
||
<pre><code>import torch.nn as nn
|
||
import torch.nn.functional as F
|
||
|
||
class Model(nn.Module):
|
||
def __init__(self):
|
||
super().__init__()
|
||
self.conv1 = nn.Conv2d(1, 20, 5)
|
||
self.conv2 = nn.Conv2d(20, 20, 5)
|
||
|
||
def forward(self, x):
|
||
x = F.relu(self.conv1(x))
|
||
return F.relu(self.conv2(x))
|
||
</code></pre>
|
||
<p>Submodules assigned in this way will be registered, and will have their
|
||
parameters converted too when you call :meth:<code>to</code>, etc.</p>
|
||
<div class="admonition note">
|
||
<p class="admonition-title">Note</p>
|
||
<p>As per the example above, an <code>__init__()</code> call to the parent class
|
||
must be made before assignment on the child.</p>
|
||
</div>
|
||
<p>:ivar training: Boolean represents whether this module is in training or
|
||
evaluation mode.
|
||
:vartype training: bool</p>
|
||
<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.models.unet.DecoderLayer.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.models.unet.DecoderLayer.dump_patches"><code class="name">var <span class="ident">dump_patches</span> : bool</code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
<dt id="audiocraft.models.unet.DecoderLayer.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.models.unet.DecoderLayer.forward"><code class="name flex">
|
||
<span>def <span class="ident">forward</span></span>(<span>self, x: torch.Tensor) ‑> torch.Tensor</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def forward(self, x: torch.Tensor) -> torch.Tensor:
|
||
x = self.res_blocks(x)
|
||
x = self.norm(x)
|
||
x = self.activation(x)
|
||
x = self.convtr(x)
|
||
return 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.models.unet.DiffusionUnet"><code class="flex name class">
|
||
<span>class <span class="ident">DiffusionUnet</span></span>
|
||
<span>(</span><span>chin: int = 3,<br>hidden: int = 24,<br>depth: int = 3,<br>growth: float = 2.0,<br>max_channels: int = 10000,<br>num_steps: int = 1000,<br>emb_all_layers=False,<br>cross_attention: bool = False,<br>bilstm: bool = False,<br>transformer: bool = False,<br>codec_dim: int | None = None,<br>**kwargs)</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">class DiffusionUnet(nn.Module):
|
||
def __init__(self, chin: int = 3, hidden: int = 24, depth: int = 3, growth: float = 2.,
|
||
max_channels: int = 10_000, num_steps: int = 1000, emb_all_layers=False, cross_attention: bool = False,
|
||
bilstm: bool = False, transformer: bool = False,
|
||
codec_dim: tp.Optional[int] = None, **kwargs):
|
||
super().__init__()
|
||
self.encoders = nn.ModuleList()
|
||
self.decoders = nn.ModuleList()
|
||
self.embeddings: tp.Optional[nn.ModuleList] = None
|
||
self.embedding = nn.Embedding(num_steps, hidden)
|
||
if emb_all_layers:
|
||
self.embeddings = nn.ModuleList()
|
||
self.condition_embedding: tp.Optional[nn.Module] = None
|
||
for d in range(depth):
|
||
encoder = EncoderLayer(chin, hidden, **kwargs)
|
||
decoder = DecoderLayer(hidden, chin, **kwargs)
|
||
self.encoders.append(encoder)
|
||
self.decoders.insert(0, decoder)
|
||
if emb_all_layers and d > 0:
|
||
assert self.embeddings is not None
|
||
self.embeddings.append(nn.Embedding(num_steps, hidden))
|
||
chin = hidden
|
||
hidden = min(int(chin * growth), max_channels)
|
||
self.bilstm: tp.Optional[nn.Module]
|
||
if bilstm:
|
||
self.bilstm = BLSTM(chin)
|
||
else:
|
||
self.bilstm = None
|
||
self.use_transformer = transformer
|
||
self.cross_attention = False
|
||
if transformer:
|
||
self.cross_attention = cross_attention
|
||
self.transformer = StreamingTransformer(chin, 8, 6, bias_ff=False, bias_attn=False,
|
||
cross_attention=cross_attention)
|
||
|
||
self.use_codec = False
|
||
if codec_dim is not None:
|
||
self.conv_codec = nn.Conv1d(codec_dim, chin, 1)
|
||
self.use_codec = True
|
||
|
||
def forward(self, x: torch.Tensor, step: tp.Union[int, torch.Tensor], condition: tp.Optional[torch.Tensor] = None):
|
||
skips = []
|
||
bs = x.size(0)
|
||
z = x
|
||
view_args = [1]
|
||
if type(step) is torch.Tensor:
|
||
step_tensor = step
|
||
else:
|
||
step_tensor = torch.tensor([step], device=x.device, dtype=torch.long).expand(bs)
|
||
|
||
for idx, encoder in enumerate(self.encoders):
|
||
z = encoder(z)
|
||
if idx == 0:
|
||
z = z + self.embedding(step_tensor).view(bs, -1, *view_args).expand_as(z)
|
||
elif self.embeddings is not None:
|
||
z = z + self.embeddings[idx - 1](step_tensor).view(bs, -1, *view_args).expand_as(z)
|
||
|
||
skips.append(z)
|
||
|
||
if self.use_codec: # insert condition in the bottleneck
|
||
assert condition is not None, "Model defined for conditionnal generation"
|
||
condition_emb = self.conv_codec(condition) # reshape to the bottleneck dim
|
||
assert condition_emb.size(-1) <= 2 * z.size(-1), \
|
||
f"You are downsampling the conditionning with factor >=2 : {condition_emb.size(-1)=} and {z.size(-1)=}"
|
||
if not self.cross_attention:
|
||
|
||
condition_emb = torch.nn.functional.interpolate(condition_emb, z.size(-1))
|
||
assert z.size() == condition_emb.size()
|
||
z += condition_emb
|
||
cross_attention_src = None
|
||
else:
|
||
cross_attention_src = condition_emb.permute(0, 2, 1) # B, T, C
|
||
B, T, C = cross_attention_src.shape
|
||
positions = torch.arange(T, device=x.device).view(1, -1, 1)
|
||
pos_emb = create_sin_embedding(positions, C, max_period=10_000, dtype=cross_attention_src.dtype)
|
||
cross_attention_src = cross_attention_src + pos_emb
|
||
if self.use_transformer:
|
||
z = self.transformer(z.permute(0, 2, 1), cross_attention_src=cross_attention_src).permute(0, 2, 1)
|
||
else:
|
||
if self.bilstm is None:
|
||
z = torch.zeros_like(z)
|
||
else:
|
||
z = self.bilstm(z)
|
||
|
||
for decoder in self.decoders:
|
||
s = skips.pop(-1)
|
||
z = z[:, :, :s.shape[2]]
|
||
z = z + s
|
||
z = decoder(z)
|
||
|
||
z = z[:, :, :x.shape[2]]
|
||
return Output(z)</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Base class for all neural network modules.</p>
|
||
<p>Your models should also subclass this class.</p>
|
||
<p>Modules can also contain other Modules, allowing to nest them in
|
||
a tree structure. You can assign the submodules as regular attributes::</p>
|
||
<pre><code>import torch.nn as nn
|
||
import torch.nn.functional as F
|
||
|
||
class Model(nn.Module):
|
||
def __init__(self):
|
||
super().__init__()
|
||
self.conv1 = nn.Conv2d(1, 20, 5)
|
||
self.conv2 = nn.Conv2d(20, 20, 5)
|
||
|
||
def forward(self, x):
|
||
x = F.relu(self.conv1(x))
|
||
return F.relu(self.conv2(x))
|
||
</code></pre>
|
||
<p>Submodules assigned in this way will be registered, and will have their
|
||
parameters converted too when you call :meth:<code>to</code>, etc.</p>
|
||
<div class="admonition note">
|
||
<p class="admonition-title">Note</p>
|
||
<p>As per the example above, an <code>__init__()</code> call to the parent class
|
||
must be made before assignment on the child.</p>
|
||
</div>
|
||
<p>:ivar training: Boolean represents whether this module is in training or
|
||
evaluation mode.
|
||
:vartype training: bool</p>
|
||
<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.models.unet.DiffusionUnet.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.models.unet.DiffusionUnet.dump_patches"><code class="name">var <span class="ident">dump_patches</span> : bool</code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
<dt id="audiocraft.models.unet.DiffusionUnet.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.models.unet.DiffusionUnet.forward"><code class="name flex">
|
||
<span>def <span class="ident">forward</span></span>(<span>self,<br>x: torch.Tensor,<br>step: int | torch.Tensor,<br>condition: torch.Tensor | None = None) ‑> Callable[..., Any]</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def forward(self, x: torch.Tensor, step: tp.Union[int, torch.Tensor], condition: tp.Optional[torch.Tensor] = None):
|
||
skips = []
|
||
bs = x.size(0)
|
||
z = x
|
||
view_args = [1]
|
||
if type(step) is torch.Tensor:
|
||
step_tensor = step
|
||
else:
|
||
step_tensor = torch.tensor([step], device=x.device, dtype=torch.long).expand(bs)
|
||
|
||
for idx, encoder in enumerate(self.encoders):
|
||
z = encoder(z)
|
||
if idx == 0:
|
||
z = z + self.embedding(step_tensor).view(bs, -1, *view_args).expand_as(z)
|
||
elif self.embeddings is not None:
|
||
z = z + self.embeddings[idx - 1](step_tensor).view(bs, -1, *view_args).expand_as(z)
|
||
|
||
skips.append(z)
|
||
|
||
if self.use_codec: # insert condition in the bottleneck
|
||
assert condition is not None, "Model defined for conditionnal generation"
|
||
condition_emb = self.conv_codec(condition) # reshape to the bottleneck dim
|
||
assert condition_emb.size(-1) <= 2 * z.size(-1), \
|
||
f"You are downsampling the conditionning with factor >=2 : {condition_emb.size(-1)=} and {z.size(-1)=}"
|
||
if not self.cross_attention:
|
||
|
||
condition_emb = torch.nn.functional.interpolate(condition_emb, z.size(-1))
|
||
assert z.size() == condition_emb.size()
|
||
z += condition_emb
|
||
cross_attention_src = None
|
||
else:
|
||
cross_attention_src = condition_emb.permute(0, 2, 1) # B, T, C
|
||
B, T, C = cross_attention_src.shape
|
||
positions = torch.arange(T, device=x.device).view(1, -1, 1)
|
||
pos_emb = create_sin_embedding(positions, C, max_period=10_000, dtype=cross_attention_src.dtype)
|
||
cross_attention_src = cross_attention_src + pos_emb
|
||
if self.use_transformer:
|
||
z = self.transformer(z.permute(0, 2, 1), cross_attention_src=cross_attention_src).permute(0, 2, 1)
|
||
else:
|
||
if self.bilstm is None:
|
||
z = torch.zeros_like(z)
|
||
else:
|
||
z = self.bilstm(z)
|
||
|
||
for decoder in self.decoders:
|
||
s = skips.pop(-1)
|
||
z = z[:, :, :s.shape[2]]
|
||
z = z + s
|
||
z = decoder(z)
|
||
|
||
z = z[:, :, :x.shape[2]]
|
||
return Output(z)</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.models.unet.EncoderLayer"><code class="flex name class">
|
||
<span>class <span class="ident">EncoderLayer</span></span>
|
||
<span>(</span><span>chin: int,<br>chout: int,<br>kernel: int = 4,<br>stride: int = 2,<br>norm_groups: int = 4,<br>res_blocks: int = 1,<br>activation: Type[torch.nn.modules.module.Module] = torch.nn.modules.activation.ReLU,<br>dropout: float = 0.0)</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">class EncoderLayer(nn.Module):
|
||
def __init__(self, chin: int, chout: int, kernel: int = 4, stride: int = 2,
|
||
norm_groups: int = 4, res_blocks: int = 1, activation: tp.Type[nn.Module] = nn.ReLU,
|
||
dropout: float = 0.):
|
||
super().__init__()
|
||
padding = (kernel - stride) // 2
|
||
Conv = nn.Conv1d
|
||
self.conv = Conv(chin, chout, kernel, stride, padding, bias=False)
|
||
self.norm = nn.GroupNorm(norm_groups, chout)
|
||
self.activation = activation()
|
||
self.res_blocks = nn.Sequential(
|
||
*[ResBlock(chout, norm_groups=norm_groups, dilation=2**idx, dropout=dropout)
|
||
for idx in range(res_blocks)])
|
||
|
||
def forward(self, x: torch.Tensor) -> torch.Tensor:
|
||
B, C, T = x.shape
|
||
stride, = self.conv.stride
|
||
pad = (stride - (T % stride)) % stride
|
||
x = F.pad(x, (0, pad))
|
||
|
||
x = self.conv(x)
|
||
x = self.norm(x)
|
||
x = self.activation(x)
|
||
x = self.res_blocks(x)
|
||
return x</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Base class for all neural network modules.</p>
|
||
<p>Your models should also subclass this class.</p>
|
||
<p>Modules can also contain other Modules, allowing to nest them in
|
||
a tree structure. You can assign the submodules as regular attributes::</p>
|
||
<pre><code>import torch.nn as nn
|
||
import torch.nn.functional as F
|
||
|
||
class Model(nn.Module):
|
||
def __init__(self):
|
||
super().__init__()
|
||
self.conv1 = nn.Conv2d(1, 20, 5)
|
||
self.conv2 = nn.Conv2d(20, 20, 5)
|
||
|
||
def forward(self, x):
|
||
x = F.relu(self.conv1(x))
|
||
return F.relu(self.conv2(x))
|
||
</code></pre>
|
||
<p>Submodules assigned in this way will be registered, and will have their
|
||
parameters converted too when you call :meth:<code>to</code>, etc.</p>
|
||
<div class="admonition note">
|
||
<p class="admonition-title">Note</p>
|
||
<p>As per the example above, an <code>__init__()</code> call to the parent class
|
||
must be made before assignment on the child.</p>
|
||
</div>
|
||
<p>:ivar training: Boolean represents whether this module is in training or
|
||
evaluation mode.
|
||
:vartype training: bool</p>
|
||
<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.models.unet.EncoderLayer.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.models.unet.EncoderLayer.dump_patches"><code class="name">var <span class="ident">dump_patches</span> : bool</code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
<dt id="audiocraft.models.unet.EncoderLayer.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.models.unet.EncoderLayer.forward"><code class="name flex">
|
||
<span>def <span class="ident">forward</span></span>(<span>self, x: torch.Tensor) ‑> torch.Tensor</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def forward(self, x: torch.Tensor) -> torch.Tensor:
|
||
B, C, T = x.shape
|
||
stride, = self.conv.stride
|
||
pad = (stride - (T % stride)) % stride
|
||
x = F.pad(x, (0, pad))
|
||
|
||
x = self.conv(x)
|
||
x = self.norm(x)
|
||
x = self.activation(x)
|
||
x = self.res_blocks(x)
|
||
return 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.models.unet.Output"><code class="flex name class">
|
||
<span>class <span class="ident">Output</span></span>
|
||
<span>(</span><span>sample: torch.Tensor)</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">@dataclass
|
||
class Output:
|
||
sample: torch.Tensor</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Output(sample: torch.Tensor)</p></div>
|
||
<h3>Class variables</h3>
|
||
<dl>
|
||
<dt id="audiocraft.models.unet.Output.sample"><code class="name">var <span class="ident">sample</span> : torch.Tensor</code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
</dl>
|
||
</dd>
|
||
<dt id="audiocraft.models.unet.ResBlock"><code class="flex name class">
|
||
<span>class <span class="ident">ResBlock</span></span>
|
||
<span>(</span><span>channels: int,<br>kernel: int = 3,<br>norm_groups: int = 4,<br>dilation: int = 1,<br>activation: Type[torch.nn.modules.module.Module] = torch.nn.modules.activation.ReLU,<br>dropout: float = 0.0)</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">class ResBlock(nn.Module):
|
||
def __init__(self, channels: int, kernel: int = 3, norm_groups: int = 4,
|
||
dilation: int = 1, activation: tp.Type[nn.Module] = nn.ReLU,
|
||
dropout: float = 0.):
|
||
super().__init__()
|
||
stride = 1
|
||
padding = dilation * (kernel - stride) // 2
|
||
Conv = nn.Conv1d
|
||
Drop = nn.Dropout1d
|
||
self.norm1 = nn.GroupNorm(norm_groups, channels)
|
||
self.conv1 = Conv(channels, channels, kernel, 1, padding, dilation=dilation)
|
||
self.activation1 = activation()
|
||
self.dropout1 = Drop(dropout)
|
||
|
||
self.norm2 = nn.GroupNorm(norm_groups, channels)
|
||
self.conv2 = Conv(channels, channels, kernel, 1, padding, dilation=dilation)
|
||
self.activation2 = activation()
|
||
self.dropout2 = Drop(dropout)
|
||
|
||
def forward(self, x):
|
||
h = self.dropout1(self.conv1(self.activation1(self.norm1(x))))
|
||
h = self.dropout2(self.conv2(self.activation2(self.norm2(h))))
|
||
return x + h</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Base class for all neural network modules.</p>
|
||
<p>Your models should also subclass this class.</p>
|
||
<p>Modules can also contain other Modules, allowing to nest them in
|
||
a tree structure. You can assign the submodules as regular attributes::</p>
|
||
<pre><code>import torch.nn as nn
|
||
import torch.nn.functional as F
|
||
|
||
class Model(nn.Module):
|
||
def __init__(self):
|
||
super().__init__()
|
||
self.conv1 = nn.Conv2d(1, 20, 5)
|
||
self.conv2 = nn.Conv2d(20, 20, 5)
|
||
|
||
def forward(self, x):
|
||
x = F.relu(self.conv1(x))
|
||
return F.relu(self.conv2(x))
|
||
</code></pre>
|
||
<p>Submodules assigned in this way will be registered, and will have their
|
||
parameters converted too when you call :meth:<code>to</code>, etc.</p>
|
||
<div class="admonition note">
|
||
<p class="admonition-title">Note</p>
|
||
<p>As per the example above, an <code>__init__()</code> call to the parent class
|
||
must be made before assignment on the child.</p>
|
||
</div>
|
||
<p>:ivar training: Boolean represents whether this module is in training or
|
||
evaluation mode.
|
||
:vartype training: bool</p>
|
||
<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.models.unet.ResBlock.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.models.unet.ResBlock.dump_patches"><code class="name">var <span class="ident">dump_patches</span> : bool</code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
<dt id="audiocraft.models.unet.ResBlock.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.models.unet.ResBlock.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):
|
||
h = self.dropout1(self.conv1(self.activation1(self.norm1(x))))
|
||
h = self.dropout2(self.conv2(self.activation2(self.norm2(h))))
|
||
return x + h</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.models" href="index.html">audiocraft.models</a></code></li>
|
||
</ul>
|
||
</li>
|
||
<li><h3><a href="#header-functions">Functions</a></h3>
|
||
<ul class="">
|
||
<li><code><a title="audiocraft.models.unet.get_model" href="#audiocraft.models.unet.get_model">get_model</a></code></li>
|
||
</ul>
|
||
</li>
|
||
<li><h3><a href="#header-classes">Classes</a></h3>
|
||
<ul>
|
||
<li>
|
||
<h4><code><a title="audiocraft.models.unet.BLSTM" href="#audiocraft.models.unet.BLSTM">BLSTM</a></code></h4>
|
||
<ul class="">
|
||
<li><code><a title="audiocraft.models.unet.BLSTM.call_super_init" href="#audiocraft.models.unet.BLSTM.call_super_init">call_super_init</a></code></li>
|
||
<li><code><a title="audiocraft.models.unet.BLSTM.dump_patches" href="#audiocraft.models.unet.BLSTM.dump_patches">dump_patches</a></code></li>
|
||
<li><code><a title="audiocraft.models.unet.BLSTM.forward" href="#audiocraft.models.unet.BLSTM.forward">forward</a></code></li>
|
||
<li><code><a title="audiocraft.models.unet.BLSTM.training" href="#audiocraft.models.unet.BLSTM.training">training</a></code></li>
|
||
</ul>
|
||
</li>
|
||
<li>
|
||
<h4><code><a title="audiocraft.models.unet.DecoderLayer" href="#audiocraft.models.unet.DecoderLayer">DecoderLayer</a></code></h4>
|
||
<ul class="">
|
||
<li><code><a title="audiocraft.models.unet.DecoderLayer.call_super_init" href="#audiocraft.models.unet.DecoderLayer.call_super_init">call_super_init</a></code></li>
|
||
<li><code><a title="audiocraft.models.unet.DecoderLayer.dump_patches" href="#audiocraft.models.unet.DecoderLayer.dump_patches">dump_patches</a></code></li>
|
||
<li><code><a title="audiocraft.models.unet.DecoderLayer.forward" href="#audiocraft.models.unet.DecoderLayer.forward">forward</a></code></li>
|
||
<li><code><a title="audiocraft.models.unet.DecoderLayer.training" href="#audiocraft.models.unet.DecoderLayer.training">training</a></code></li>
|
||
</ul>
|
||
</li>
|
||
<li>
|
||
<h4><code><a title="audiocraft.models.unet.DiffusionUnet" href="#audiocraft.models.unet.DiffusionUnet">DiffusionUnet</a></code></h4>
|
||
<ul class="">
|
||
<li><code><a title="audiocraft.models.unet.DiffusionUnet.call_super_init" href="#audiocraft.models.unet.DiffusionUnet.call_super_init">call_super_init</a></code></li>
|
||
<li><code><a title="audiocraft.models.unet.DiffusionUnet.dump_patches" href="#audiocraft.models.unet.DiffusionUnet.dump_patches">dump_patches</a></code></li>
|
||
<li><code><a title="audiocraft.models.unet.DiffusionUnet.forward" href="#audiocraft.models.unet.DiffusionUnet.forward">forward</a></code></li>
|
||
<li><code><a title="audiocraft.models.unet.DiffusionUnet.training" href="#audiocraft.models.unet.DiffusionUnet.training">training</a></code></li>
|
||
</ul>
|
||
</li>
|
||
<li>
|
||
<h4><code><a title="audiocraft.models.unet.EncoderLayer" href="#audiocraft.models.unet.EncoderLayer">EncoderLayer</a></code></h4>
|
||
<ul class="">
|
||
<li><code><a title="audiocraft.models.unet.EncoderLayer.call_super_init" href="#audiocraft.models.unet.EncoderLayer.call_super_init">call_super_init</a></code></li>
|
||
<li><code><a title="audiocraft.models.unet.EncoderLayer.dump_patches" href="#audiocraft.models.unet.EncoderLayer.dump_patches">dump_patches</a></code></li>
|
||
<li><code><a title="audiocraft.models.unet.EncoderLayer.forward" href="#audiocraft.models.unet.EncoderLayer.forward">forward</a></code></li>
|
||
<li><code><a title="audiocraft.models.unet.EncoderLayer.training" href="#audiocraft.models.unet.EncoderLayer.training">training</a></code></li>
|
||
</ul>
|
||
</li>
|
||
<li>
|
||
<h4><code><a title="audiocraft.models.unet.Output" href="#audiocraft.models.unet.Output">Output</a></code></h4>
|
||
<ul class="">
|
||
<li><code><a title="audiocraft.models.unet.Output.sample" href="#audiocraft.models.unet.Output.sample">sample</a></code></li>
|
||
</ul>
|
||
</li>
|
||
<li>
|
||
<h4><code><a title="audiocraft.models.unet.ResBlock" href="#audiocraft.models.unet.ResBlock">ResBlock</a></code></h4>
|
||
<ul class="">
|
||
<li><code><a title="audiocraft.models.unet.ResBlock.call_super_init" href="#audiocraft.models.unet.ResBlock.call_super_init">call_super_init</a></code></li>
|
||
<li><code><a title="audiocraft.models.unet.ResBlock.dump_patches" href="#audiocraft.models.unet.ResBlock.dump_patches">dump_patches</a></code></li>
|
||
<li><code><a title="audiocraft.models.unet.ResBlock.forward" href="#audiocraft.models.unet.ResBlock.forward">forward</a></code></li>
|
||
<li><code><a title="audiocraft.models.unet.ResBlock.training" href="#audiocraft.models.unet.ResBlock.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>
|