facebookresearch--audiocraft
Alexandre Défossez (autodoc)
673 行
36 KiB
HTML
673 行
36 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.solvers.builders API documentation</title>
|
||
<meta name="description" content="All the functions to build the relevant solvers and used objects
|
||
from the Hydra config.">
|
||
<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.solvers.builders</code></h1>
|
||
</header>
|
||
<section id="section-intro">
|
||
<p>All the functions to build the relevant solvers and used objects
|
||
from the Hydra config.</p>
|
||
</section>
|
||
<section>
|
||
</section>
|
||
<section>
|
||
</section>
|
||
<section>
|
||
<h2 class="section-title" id="header-functions">Functions</h2>
|
||
<dl>
|
||
<dt id="audiocraft.solvers.builders.get_adversarial_losses"><code class="name flex">
|
||
<span>def <span class="ident">get_adversarial_losses</span></span>(<span>cfg) ‑> torch.nn.modules.container.ModuleDict</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_adversarial_losses(cfg) -> nn.ModuleDict:
|
||
"""Initialize dict of adversarial losses from config."""
|
||
device = cfg.device
|
||
adv_cfg = getattr(cfg, 'adversarial')
|
||
adversaries = adv_cfg.get('adversaries', [])
|
||
adv_loss_name = adv_cfg['adv_loss']
|
||
feat_loss_name = adv_cfg.get('feat_loss')
|
||
normalize = adv_cfg.get('normalize', True)
|
||
feat_loss: tp.Optional[adversarial.FeatureMatchingLoss] = None
|
||
if feat_loss_name:
|
||
assert feat_loss_name in ['l1', 'l2'], f"Feature loss only support L1 or L2 but {feat_loss_name} found."
|
||
loss = get_loss(feat_loss_name, cfg)
|
||
feat_loss = adversarial.FeatureMatchingLoss(loss, normalize)
|
||
loss = adversarial.get_adv_criterion(adv_loss_name)
|
||
loss_real = adversarial.get_real_criterion(adv_loss_name)
|
||
loss_fake = adversarial.get_fake_criterion(adv_loss_name)
|
||
adv_losses = nn.ModuleDict()
|
||
for adv_name in adversaries:
|
||
adversary = get_adversary(adv_name, cfg).to(device)
|
||
optimizer = get_optimizer(adversary.parameters(), cfg.optim)
|
||
adv_loss = adversarial.AdversarialLoss(
|
||
adversary,
|
||
optimizer,
|
||
loss=loss,
|
||
loss_real=loss_real,
|
||
loss_fake=loss_fake,
|
||
loss_feat=feat_loss,
|
||
normalize=normalize
|
||
)
|
||
adv_losses[adv_name] = adv_loss
|
||
return adv_losses</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Initialize dict of adversarial losses from config.</p></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.get_adversary"><code class="name flex">
|
||
<span>def <span class="ident">get_adversary</span></span>(<span>name: str, cfg: omegaconf.dictconfig.DictConfig) ‑> torch.nn.modules.module.Module</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_adversary(name: str, cfg: omegaconf.DictConfig) -> nn.Module:
|
||
"""Initialize adversary from config."""
|
||
klass = {
|
||
'msd': adversarial.MultiScaleDiscriminator,
|
||
'mpd': adversarial.MultiPeriodDiscriminator,
|
||
'msstftd': adversarial.MultiScaleSTFTDiscriminator,
|
||
}[name]
|
||
adv_cfg: tp.Dict[str, tp.Any] = dict(getattr(cfg, name))
|
||
return klass(**adv_cfg)</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Initialize adversary from config.</p></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.get_audio_datasets"><code class="name flex">
|
||
<span>def <span class="ident">get_audio_datasets</span></span>(<span>cfg: omegaconf.dictconfig.DictConfig,<br>dataset_type: <a title="audiocraft.solvers.builders.DatasetType" href="#audiocraft.solvers.builders.DatasetType">DatasetType</a> = DatasetType.AUDIO) ‑> Dict[str, torch.utils.data.dataloader.DataLoader]</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_audio_datasets(cfg: omegaconf.DictConfig,
|
||
dataset_type: DatasetType = DatasetType.AUDIO) -> tp.Dict[str, torch.utils.data.DataLoader]:
|
||
"""Build AudioDataset from configuration.
|
||
|
||
Args:
|
||
cfg (omegaconf.DictConfig): Configuration.
|
||
dataset_type: The type of dataset to create.
|
||
Returns:
|
||
dict[str, torch.utils.data.DataLoader]: Map of dataloader for each data split.
|
||
"""
|
||
dataloaders: dict = {}
|
||
|
||
sample_rate = cfg.sample_rate
|
||
channels = cfg.channels
|
||
seed = cfg.seed
|
||
max_sample_rate = cfg.datasource.max_sample_rate
|
||
max_channels = cfg.datasource.max_channels
|
||
|
||
assert cfg.dataset is not None, "Could not find dataset definition in config"
|
||
|
||
dataset_cfg = dict_from_config(cfg.dataset)
|
||
splits_cfg: dict = {}
|
||
splits_cfg['train'] = dataset_cfg.pop('train')
|
||
splits_cfg['valid'] = dataset_cfg.pop('valid')
|
||
splits_cfg['evaluate'] = dataset_cfg.pop('evaluate')
|
||
splits_cfg['generate'] = dataset_cfg.pop('generate')
|
||
execute_only_stage = cfg.get('execute_only', None)
|
||
|
||
for split, path in cfg.datasource.items():
|
||
if not isinstance(path, str):
|
||
continue # skipping this as not a path
|
||
if execute_only_stage is not None and split != execute_only_stage:
|
||
continue
|
||
logger.info(f"Loading audio data split {split}: {str(path)}")
|
||
assert (
|
||
cfg.sample_rate <= max_sample_rate
|
||
), f"Expecting a max sample rate of {max_sample_rate} for datasource but {sample_rate} found."
|
||
assert (
|
||
cfg.channels <= max_channels
|
||
), f"Expecting a max number of channels of {max_channels} for datasource but {channels} found."
|
||
|
||
split_cfg = splits_cfg[split]
|
||
split_kwargs = {k: v for k, v in split_cfg.items()}
|
||
kwargs = {**dataset_cfg, **split_kwargs} # split kwargs overrides default dataset_cfg
|
||
kwargs['sample_rate'] = sample_rate
|
||
kwargs['channels'] = channels
|
||
|
||
if kwargs.get('permutation_on_files') and cfg.optim.updates_per_epoch:
|
||
kwargs['num_samples'] = (
|
||
flashy.distrib.world_size() * cfg.dataset.batch_size * cfg.optim.updates_per_epoch)
|
||
|
||
num_samples = kwargs['num_samples']
|
||
shuffle = kwargs['shuffle']
|
||
|
||
return_info = kwargs.pop('return_info')
|
||
batch_size = kwargs.pop('batch_size', None)
|
||
num_workers = kwargs.pop('num_workers')
|
||
|
||
if dataset_type == DatasetType.MUSIC:
|
||
dataset = data.music_dataset.MusicDataset.from_meta(path, **kwargs)
|
||
elif dataset_type == DatasetType.SOUND:
|
||
dataset = data.sound_dataset.SoundDataset.from_meta(path, **kwargs)
|
||
elif dataset_type == DatasetType.AUDIO:
|
||
dataset = data.info_audio_dataset.InfoAudioDataset.from_meta(path, return_info=return_info, **kwargs)
|
||
elif dataset_type == DatasetType.JASCO:
|
||
dataset = data.jasco_dataset.JascoDataset.from_meta(path, return_info=return_info, **kwargs)
|
||
else:
|
||
raise ValueError(f"Dataset type is unsupported: {dataset_type}")
|
||
|
||
loader = get_loader(
|
||
dataset,
|
||
num_samples,
|
||
batch_size=batch_size,
|
||
num_workers=num_workers,
|
||
seed=seed,
|
||
collate_fn=dataset.collater if return_info else None,
|
||
shuffle=shuffle,
|
||
)
|
||
dataloaders[split] = loader
|
||
|
||
return dataloaders</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Build AudioDataset from configuration.</p>
|
||
<h2 id="args">Args</h2>
|
||
<dl>
|
||
<dt><strong><code>cfg</code></strong> : <code>omegaconf.DictConfig</code></dt>
|
||
<dd>Configuration.</dd>
|
||
<dt><strong><code>dataset_type</code></strong></dt>
|
||
<dd>The type of dataset to create.</dd>
|
||
</dl>
|
||
<h2 id="returns">Returns</h2>
|
||
<dl>
|
||
<dt><code>dict[str, torch.utils.data.DataLoader]</code></dt>
|
||
<dd>Map of dataloader for each data split.</dd>
|
||
</dl></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.get_balancer"><code class="name flex">
|
||
<span>def <span class="ident">get_balancer</span></span>(<span>loss_weights: Dict[str, float], cfg: omegaconf.dictconfig.DictConfig) ‑> <a title="audiocraft.losses.balancer.Balancer" href="../losses/balancer.html#audiocraft.losses.balancer.Balancer">Balancer</a></span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_balancer(loss_weights: tp.Dict[str, float], cfg: omegaconf.DictConfig) -> losses.Balancer:
|
||
"""Instantiate loss balancer from configuration for the provided weights."""
|
||
kwargs: tp.Dict[str, tp.Any] = dict_from_config(cfg)
|
||
return losses.Balancer(loss_weights, **kwargs)</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Instantiate loss balancer from configuration for the provided weights.</p></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.get_chroma_cosine_similarity"><code class="name flex">
|
||
<span>def <span class="ident">get_chroma_cosine_similarity</span></span>(<span>cfg: omegaconf.dictconfig.DictConfig) ‑> <a title="audiocraft.metrics.chroma_cosinesim.ChromaCosineSimilarityMetric" href="../metrics/chroma_cosinesim.html#audiocraft.metrics.chroma_cosinesim.ChromaCosineSimilarityMetric">ChromaCosineSimilarityMetric</a></span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_chroma_cosine_similarity(cfg: omegaconf.DictConfig) -> metrics.ChromaCosineSimilarityMetric:
|
||
"""Instantiate Chroma Cosine Similarity metric from config."""
|
||
assert cfg.model == 'chroma_base', "Only support 'chroma_base' method for chroma cosine similarity metric"
|
||
kwargs = dict_from_config(cfg.get(cfg.model))
|
||
return metrics.ChromaCosineSimilarityMetric(**kwargs)</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Instantiate Chroma Cosine Similarity metric from config.</p></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.get_ema"><code class="name flex">
|
||
<span>def <span class="ident">get_ema</span></span>(<span>module_dict: torch.nn.modules.container.ModuleDict,<br>cfg: omegaconf.dictconfig.DictConfig) ‑> <a title="audiocraft.optim.ema.ModuleDictEMA" href="../optim/ema.html#audiocraft.optim.ema.ModuleDictEMA">ModuleDictEMA</a> | None</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_ema(module_dict: nn.ModuleDict, cfg: omegaconf.DictConfig) -> tp.Optional[optim.ModuleDictEMA]:
|
||
"""Initialize Exponential Moving Average.
|
||
|
||
Args:
|
||
module_dict (nn.ModuleDict): ModuleDict for which to compute the EMA.
|
||
cfg (omegaconf.DictConfig): Optim EMA configuration.
|
||
Returns:
|
||
optim.ModuleDictEMA: EMA version of the ModuleDict.
|
||
"""
|
||
kw: tp.Dict[str, tp.Any] = dict(cfg)
|
||
use = kw.pop('use', False)
|
||
decay = kw.pop('decay', None)
|
||
device = kw.pop('device', None)
|
||
if not use:
|
||
return None
|
||
if len(module_dict) == 0:
|
||
raise ValueError("Trying to build EMA but an empty module_dict source is provided!")
|
||
ema_module = optim.ModuleDictEMA(module_dict, decay=decay, device=device)
|
||
return ema_module</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Initialize Exponential Moving Average.</p>
|
||
<h2 id="args">Args</h2>
|
||
<dl>
|
||
<dt><strong><code>module_dict</code></strong> : <code>nn.ModuleDict</code></dt>
|
||
<dd>ModuleDict for which to compute the EMA.</dd>
|
||
<dt><strong><code>cfg</code></strong> : <code>omegaconf.DictConfig</code></dt>
|
||
<dd>Optim EMA configuration.</dd>
|
||
</dl>
|
||
<h2 id="returns">Returns</h2>
|
||
<dl>
|
||
<dt><code>optim.ModuleDictEMA</code></dt>
|
||
<dd>EMA version of the ModuleDict.</dd>
|
||
</dl></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.get_fad"><code class="name flex">
|
||
<span>def <span class="ident">get_fad</span></span>(<span>cfg: omegaconf.dictconfig.DictConfig) ‑> <a title="audiocraft.metrics.fad.FrechetAudioDistanceMetric" href="../metrics/fad.html#audiocraft.metrics.fad.FrechetAudioDistanceMetric">FrechetAudioDistanceMetric</a></span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_fad(cfg: omegaconf.DictConfig) -> metrics.FrechetAudioDistanceMetric:
|
||
"""Instantiate Frechet Audio Distance metric from config."""
|
||
kwargs = dict_from_config(cfg.tf)
|
||
xp = dora.get_xp()
|
||
kwargs['log_folder'] = xp.folder
|
||
return metrics.FrechetAudioDistanceMetric(**kwargs)</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Instantiate Frechet Audio Distance metric from config.</p></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.get_kldiv"><code class="name flex">
|
||
<span>def <span class="ident">get_kldiv</span></span>(<span>cfg: omegaconf.dictconfig.DictConfig) ‑> <a title="audiocraft.metrics.kld.KLDivergenceMetric" href="../metrics/kld.html#audiocraft.metrics.kld.KLDivergenceMetric">KLDivergenceMetric</a></span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_kldiv(cfg: omegaconf.DictConfig) -> metrics.KLDivergenceMetric:
|
||
"""Instantiate KL-Divergence metric from config."""
|
||
kld_metrics = {
|
||
'passt': metrics.PasstKLDivergenceMetric,
|
||
}
|
||
klass = kld_metrics[cfg.model]
|
||
kwargs = dict_from_config(cfg.get(cfg.model))
|
||
return klass(**kwargs)</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Instantiate KL-Divergence metric from config.</p></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.get_loss"><code class="name flex">
|
||
<span>def <span class="ident">get_loss</span></span>(<span>loss_name: str, cfg: omegaconf.dictconfig.DictConfig)</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_loss(loss_name: str, cfg: omegaconf.DictConfig):
|
||
"""Instantiate loss from configuration."""
|
||
klass = {
|
||
'l1': torch.nn.L1Loss,
|
||
'l2': torch.nn.MSELoss,
|
||
'mel': losses.MelSpectrogramL1Loss,
|
||
'mrstft': losses.MRSTFTLoss,
|
||
'msspec': losses.MultiScaleMelSpectrogramLoss,
|
||
'sisnr': losses.SISNR,
|
||
'wm_detection': losses.WMDetectionLoss,
|
||
'wm_mb': losses.WMMbLoss,
|
||
'tf_loudnessratio': losses.TFLoudnessRatio
|
||
}[loss_name]
|
||
kwargs = dict(getattr(cfg, loss_name))
|
||
return klass(**kwargs)</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Instantiate loss from configuration.</p></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.get_lr_scheduler"><code class="name flex">
|
||
<span>def <span class="ident">get_lr_scheduler</span></span>(<span>optimizer: torch.optim.optimizer.Optimizer,<br>cfg: omegaconf.dictconfig.DictConfig,<br>total_updates: int) ‑> torch.optim.lr_scheduler.LRScheduler | None</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_lr_scheduler(optimizer: torch.optim.Optimizer,
|
||
cfg: omegaconf.DictConfig,
|
||
total_updates: int) -> tp.Optional[LRScheduler]:
|
||
"""Build torch learning rate scheduler from config and associated optimizer.
|
||
Supported learning rate schedulers: ExponentialLRScheduler, PlateauLRScheduler
|
||
|
||
Args:
|
||
optimizer (torch.optim.Optimizer): Optimizer.
|
||
cfg (DictConfig): Schedule-related configuration.
|
||
total_updates (int): Total number of updates.
|
||
Returns:
|
||
torch.optim.Optimizer.
|
||
"""
|
||
if 'lr_scheduler' not in cfg:
|
||
raise KeyError("LR Scheduler not found in config")
|
||
|
||
lr_sched: tp.Optional[LRScheduler] = None
|
||
if cfg.lr_scheduler == 'step':
|
||
lr_sched = torch.optim.lr_scheduler.StepLR(optimizer, **cfg.step)
|
||
elif cfg.lr_scheduler == 'exponential':
|
||
lr_sched = torch.optim.lr_scheduler.ExponentialLR(optimizer, gamma=cfg.exponential)
|
||
elif cfg.lr_scheduler == 'cosine':
|
||
kwargs = dict_from_config(cfg.cosine)
|
||
warmup_steps = kwargs.pop('warmup')
|
||
lr_sched = optim.CosineLRScheduler(
|
||
optimizer, warmup_steps=warmup_steps, total_steps=total_updates, **kwargs)
|
||
elif cfg.lr_scheduler == 'polynomial_decay':
|
||
kwargs = dict_from_config(cfg.polynomial_decay)
|
||
warmup_steps = kwargs.pop('warmup')
|
||
lr_sched = optim.PolynomialDecayLRScheduler(
|
||
optimizer, warmup_steps=warmup_steps, total_steps=total_updates, **kwargs)
|
||
elif cfg.lr_scheduler == 'inverse_sqrt':
|
||
kwargs = dict_from_config(cfg.inverse_sqrt)
|
||
warmup_steps = kwargs.pop('warmup')
|
||
lr_sched = optim.InverseSquareRootLRScheduler(optimizer, warmup_steps=warmup_steps, **kwargs)
|
||
elif cfg.lr_scheduler == 'linear_warmup':
|
||
kwargs = dict_from_config(cfg.linear_warmup)
|
||
warmup_steps = kwargs.pop('warmup')
|
||
lr_sched = optim.LinearWarmupLRScheduler(optimizer, warmup_steps=warmup_steps, **kwargs)
|
||
elif cfg.lr_scheduler is not None:
|
||
raise ValueError(f"Unsupported LR Scheduler: {cfg.lr_scheduler}")
|
||
return lr_sched</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Build torch learning rate scheduler from config and associated optimizer.
|
||
Supported learning rate schedulers: ExponentialLRScheduler, PlateauLRScheduler</p>
|
||
<h2 id="args">Args</h2>
|
||
<dl>
|
||
<dt><strong><code>optimizer</code></strong> : <code>torch.optim.Optimizer</code></dt>
|
||
<dd>Optimizer.</dd>
|
||
<dt><strong><code>cfg</code></strong> : <code>DictConfig</code></dt>
|
||
<dd>Schedule-related configuration.</dd>
|
||
<dt><strong><code>total_updates</code></strong> : <code>int</code></dt>
|
||
<dd>Total number of updates.</dd>
|
||
</dl>
|
||
<h2 id="returns">Returns</h2>
|
||
<p>torch.optim.Optimizer.</p></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.get_optim_parameter_groups"><code class="name flex">
|
||
<span>def <span class="ident">get_optim_parameter_groups</span></span>(<span>model: torch.nn.modules.module.Module)</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_optim_parameter_groups(model: nn.Module):
|
||
"""Create parameter groups for the model using the appropriate method
|
||
if defined for each modules, to create the different groups.
|
||
|
||
Args:
|
||
model (nn.Module): torch model
|
||
Returns:
|
||
List of parameter groups
|
||
"""
|
||
seen_params: tp.Set[nn.parameter.Parameter] = set()
|
||
other_params = []
|
||
groups = []
|
||
for name, module in model.named_modules():
|
||
if hasattr(module, 'make_optim_group'):
|
||
group = module.make_optim_group()
|
||
params = set(group['params'])
|
||
assert params.isdisjoint(seen_params)
|
||
seen_params |= set(params)
|
||
groups.append(group)
|
||
for param in model.parameters():
|
||
if param not in seen_params:
|
||
other_params.append(param)
|
||
groups.insert(0, {'params': other_params})
|
||
parameters = groups
|
||
return parameters</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Create parameter groups for the model using the appropriate method
|
||
if defined for each modules, to create the different groups.</p>
|
||
<h2 id="args">Args</h2>
|
||
<dl>
|
||
<dt><strong><code>model</code></strong> : <code>nn.Module</code></dt>
|
||
<dd>torch model</dd>
|
||
</dl>
|
||
<h2 id="returns">Returns</h2>
|
||
<p>List of parameter groups</p></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.get_optimizer"><code class="name flex">
|
||
<span>def <span class="ident">get_optimizer</span></span>(<span>params: torch.nn.modules.module.Module | Iterable[torch.Tensor],<br>cfg: omegaconf.dictconfig.DictConfig) ‑> torch.optim.optimizer.Optimizer</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_optimizer(params: tp.Union[nn.Module, tp.Iterable[torch.Tensor]], cfg: omegaconf.DictConfig) -> Optimizer:
|
||
"""Build torch optimizer from config and set of parameters.
|
||
Supported optimizers: Adam, AdamW
|
||
|
||
Args:
|
||
params (nn.Module or iterable of torch.Tensor): Parameters to optimize.
|
||
cfg (DictConfig): Optimization-related configuration.
|
||
Returns:
|
||
torch.optim.Optimizer.
|
||
"""
|
||
if 'optimizer' not in cfg:
|
||
if getattr(cfg, 'optim', None) is not None:
|
||
raise KeyError("Optimizer not found in config. Try instantiating optimizer from cfg.optim?")
|
||
else:
|
||
raise KeyError("Optimizer not found in config.")
|
||
|
||
parameters = get_optim_parameter_groups(params) if isinstance(params, nn.Module) else params
|
||
optimizer: torch.optim.Optimizer
|
||
if cfg.optimizer == 'adam':
|
||
optimizer = torch.optim.Adam(parameters, lr=cfg.lr, **cfg.adam)
|
||
elif cfg.optimizer == 'adamw':
|
||
optimizer = torch.optim.AdamW(parameters, lr=cfg.lr, **cfg.adam)
|
||
elif cfg.optimizer == 'dadam':
|
||
optimizer = optim.DAdaptAdam(parameters, lr=cfg.lr, **cfg.adam)
|
||
else:
|
||
raise ValueError(f"Unsupported Optimizer: {cfg.optimizer}")
|
||
return optimizer</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Build torch optimizer from config and set of parameters.
|
||
Supported optimizers: Adam, AdamW</p>
|
||
<h2 id="args">Args</h2>
|
||
<dl>
|
||
<dt><strong><code>params</code></strong> : <code>nn.Module</code> or <code>iterable</code> of <code>torch.Tensor</code></dt>
|
||
<dd>Parameters to optimize.</dd>
|
||
<dt><strong><code>cfg</code></strong> : <code>DictConfig</code></dt>
|
||
<dd>Optimization-related configuration.</dd>
|
||
</dl>
|
||
<h2 id="returns">Returns</h2>
|
||
<p>torch.optim.Optimizer.</p></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.get_solver"><code class="name flex">
|
||
<span>def <span class="ident">get_solver</span></span>(<span>cfg: omegaconf.dictconfig.DictConfig) ‑> <a title="audiocraft.solvers.base.StandardSolver" href="base.html#audiocraft.solvers.base.StandardSolver">StandardSolver</a></span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_solver(cfg: omegaconf.DictConfig) -> StandardSolver:
|
||
"""Instantiate solver from config."""
|
||
from .audiogen import AudioGenSolver
|
||
from .compression import CompressionSolver
|
||
from .musicgen import MusicGenSolver
|
||
from .diffusion import DiffusionSolver
|
||
from .magnet import MagnetSolver, AudioMagnetSolver
|
||
from .watermark import WatermarkSolver
|
||
from .jasco import JascoSolver
|
||
klass = {
|
||
'compression': CompressionSolver,
|
||
'musicgen': MusicGenSolver,
|
||
'audiogen': AudioGenSolver,
|
||
'magnet': MagnetSolver,
|
||
'audio_magnet': AudioMagnetSolver,
|
||
'lm': MusicGenSolver, # backward compatibility
|
||
'diffusion': DiffusionSolver,
|
||
'sound_lm': AudioGenSolver, # backward compatibility
|
||
'watermarking': WatermarkSolver,
|
||
'jasco': JascoSolver,
|
||
}[cfg.solver]
|
||
return klass(cfg) # type: ignore</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Instantiate solver from config.</p></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.get_text_consistency"><code class="name flex">
|
||
<span>def <span class="ident">get_text_consistency</span></span>(<span>cfg: omegaconf.dictconfig.DictConfig) ‑> <a title="audiocraft.metrics.clap_consistency.TextConsistencyMetric" href="../metrics/clap_consistency.html#audiocraft.metrics.clap_consistency.TextConsistencyMetric">TextConsistencyMetric</a></span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_text_consistency(cfg: omegaconf.DictConfig) -> metrics.TextConsistencyMetric:
|
||
"""Instantiate Text Consistency metric from config."""
|
||
text_consistency_metrics = {
|
||
'clap': metrics.CLAPTextConsistencyMetric
|
||
}
|
||
klass = text_consistency_metrics[cfg.model]
|
||
kwargs = dict_from_config(cfg.get(cfg.model))
|
||
return klass(**kwargs)</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Instantiate Text Consistency metric from config.</p></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.get_visqol"><code class="name flex">
|
||
<span>def <span class="ident">get_visqol</span></span>(<span>cfg: omegaconf.dictconfig.DictConfig) ‑> <a title="audiocraft.metrics.visqol.ViSQOL" href="../metrics/visqol.html#audiocraft.metrics.visqol.ViSQOL">ViSQOL</a></span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">def get_visqol(cfg: omegaconf.DictConfig) -> metrics.ViSQOL:
|
||
"""Instantiate ViSQOL metric from config."""
|
||
kwargs = dict_from_config(cfg)
|
||
return metrics.ViSQOL(**kwargs)</code></pre>
|
||
</details>
|
||
<div class="desc"><p>Instantiate ViSQOL metric from config.</p></div>
|
||
</dd>
|
||
</dl>
|
||
</section>
|
||
<section>
|
||
<h2 class="section-title" id="header-classes">Classes</h2>
|
||
<dl>
|
||
<dt id="audiocraft.solvers.builders.DatasetType"><code class="flex name class">
|
||
<span>class <span class="ident">DatasetType</span></span>
|
||
<span>(</span><span>value, names=None, *, module=None, qualname=None, type=None, start=1)</span>
|
||
</code></dt>
|
||
<dd>
|
||
<details class="source">
|
||
<summary>
|
||
<span>Expand source code</span>
|
||
</summary>
|
||
<pre><code class="python">class DatasetType(Enum):
|
||
AUDIO = "audio"
|
||
MUSIC = "music"
|
||
SOUND = "sound"
|
||
JASCO = "jasco"</code></pre>
|
||
</details>
|
||
<div class="desc"><p>An enumeration.</p></div>
|
||
<h3>Ancestors</h3>
|
||
<ul class="hlist">
|
||
<li>enum.Enum</li>
|
||
</ul>
|
||
<h3>Class variables</h3>
|
||
<dl>
|
||
<dt id="audiocraft.solvers.builders.DatasetType.AUDIO"><code class="name">var <span class="ident">AUDIO</span></code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.DatasetType.JASCO"><code class="name">var <span class="ident">JASCO</span></code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.DatasetType.MUSIC"><code class="name">var <span class="ident">MUSIC</span></code></dt>
|
||
<dd>
|
||
<div class="desc"></div>
|
||
</dd>
|
||
<dt id="audiocraft.solvers.builders.DatasetType.SOUND"><code class="name">var <span class="ident">SOUND</span></code></dt>
|
||
<dd>
|
||
<div class="desc"></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.solvers" href="index.html">audiocraft.solvers</a></code></li>
|
||
</ul>
|
||
</li>
|
||
<li><h3><a href="#header-functions">Functions</a></h3>
|
||
<ul class="">
|
||
<li><code><a title="audiocraft.solvers.builders.get_adversarial_losses" href="#audiocraft.solvers.builders.get_adversarial_losses">get_adversarial_losses</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.get_adversary" href="#audiocraft.solvers.builders.get_adversary">get_adversary</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.get_audio_datasets" href="#audiocraft.solvers.builders.get_audio_datasets">get_audio_datasets</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.get_balancer" href="#audiocraft.solvers.builders.get_balancer">get_balancer</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.get_chroma_cosine_similarity" href="#audiocraft.solvers.builders.get_chroma_cosine_similarity">get_chroma_cosine_similarity</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.get_ema" href="#audiocraft.solvers.builders.get_ema">get_ema</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.get_fad" href="#audiocraft.solvers.builders.get_fad">get_fad</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.get_kldiv" href="#audiocraft.solvers.builders.get_kldiv">get_kldiv</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.get_loss" href="#audiocraft.solvers.builders.get_loss">get_loss</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.get_lr_scheduler" href="#audiocraft.solvers.builders.get_lr_scheduler">get_lr_scheduler</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.get_optim_parameter_groups" href="#audiocraft.solvers.builders.get_optim_parameter_groups">get_optim_parameter_groups</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.get_optimizer" href="#audiocraft.solvers.builders.get_optimizer">get_optimizer</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.get_solver" href="#audiocraft.solvers.builders.get_solver">get_solver</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.get_text_consistency" href="#audiocraft.solvers.builders.get_text_consistency">get_text_consistency</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.get_visqol" href="#audiocraft.solvers.builders.get_visqol">get_visqol</a></code></li>
|
||
</ul>
|
||
</li>
|
||
<li><h3><a href="#header-classes">Classes</a></h3>
|
||
<ul>
|
||
<li>
|
||
<h4><code><a title="audiocraft.solvers.builders.DatasetType" href="#audiocraft.solvers.builders.DatasetType">DatasetType</a></code></h4>
|
||
<ul class="">
|
||
<li><code><a title="audiocraft.solvers.builders.DatasetType.AUDIO" href="#audiocraft.solvers.builders.DatasetType.AUDIO">AUDIO</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.DatasetType.JASCO" href="#audiocraft.solvers.builders.DatasetType.JASCO">JASCO</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.DatasetType.MUSIC" href="#audiocraft.solvers.builders.DatasetType.MUSIC">MUSIC</a></code></li>
|
||
<li><code><a title="audiocraft.solvers.builders.DatasetType.SOUND" href="#audiocraft.solvers.builders.DatasetType.SOUND">SOUND</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>
|