mirrors / facebookresearch--audiocraft

项目文件夹

facebookresearch--audiocraft
文件
gh-docs
T
Alexandre Défossez (autodoc) aeffa9f270 api_docs
2025-03-13 16:07:03 +00:00

604 行
35 KiB
HTML
原始文件 永久链接 Blame 文件历史

此文件含有不可见的 Unicode 字符
此文件含有人类无法区分的不可见的 Unicode 字符，但可以由计算机进行不同的处理。 如果您是想特意这样的，可以安全地忽略该警告。 使用 Escape 按钮显示他们。
此文件含有可能会与其他字符混淆的 Unicode 字符。 如果您是想特意这样的，可以安全地忽略该警告。 使用 Escape 按钮显示他们。
<!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.jasco API documentation</title>
<meta name="description" content="Main model for using JASCO. This will combine all the required components
and provide easy access to the generation API.">
<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.jasco</code></h1>
</header>
<section id="section-intro">
<p>Main model for using JASCO. This will combine all the required components
and provide easy access to the generation API.</p>
</section>
<section>
</section>
<section>
</section>
<section>
</section>
<section>
<h2 class="section-title" id="header-classes">Classes</h2>
<dl>
<dt id="audiocraft.models.jasco.JASCO"><code class="flex name class">
<span>class <span class="ident">JASCO</span></span>
<span>(</span><span>chords_mapping_path='assets/chord_to_index_mapping.pkl', **kwargs)</span>
</code></dt>
<dd>
<details class="source">
<summary>
<span>Expand source code</span>
</summary>
<pre><code class="python">class JASCO(BaseGenModel):
&#34;&#34;&#34;JASCO main model with convenient generation API.
Args:
chords_mapping_path: path to chords to index mapping pickle
kwargs - See MusicGen class.
&#34;&#34;&#34;
def __init__(self, chords_mapping_path=&#39;assets/chord_to_index_mapping.pkl&#39;, **kwargs):
super().__init__(**kwargs)
# JASCO operates over a fixed sequence length defined in it&#39;s config.
self.duration = self.lm.cfg.dataset.segment_duration
# load chord2index mapping of Chordino (https://github.com/ohollo/chord-extractor)
assert os.path.exists(chords_mapping_path)
self.chords_mapping = pickle.load(open(chords_mapping_path, &#34;rb&#34;))
# set generation parameters
self.set_generation_params()
@staticmethod
def get_pretrained(name: str = &#39;facebook/jasco-chords-drums-400M&#39;, device=None,
chords_mapping_path=&#39;assets/chord_to_index_mapping.pkl&#39;):
&#34;&#34;&#34;Return pretrained model, we provide 2 models:
1. facebook/jasco-chords-drums-400M: 10s music generation conditioned on
text, chords and drums, 400M parameters.
2. facebook/jasco-chords-drums-1B: 10s music generation conditioned on
text, chords and drums, 1B parameters.
&#34;&#34;&#34;
if device is None:
if torch.cuda.device_count():
device = &#39;cuda&#39;
else:
device = &#39;cpu&#39;
compression_model = load_compression_model(name, device=device)
lm = load_jasco_model(name, compression_model, device=device)
kwargs = {&#39;name&#39;: name,
&#39;compression_model&#39;: compression_model,
&#39;lm&#39;: lm,
&#39;chords_mapping_path&#39;: chords_mapping_path}
return JASCO(**kwargs)
def set_generation_params(self,
cfg_coef_all: float = 5.0,
cfg_coef_txt: float = 0.0,
**kwargs):
&#34;&#34;&#34;Set the generation parameters for JASCO.
Args:
cfg_coef_all (float, optional): Coefficient used in multi-source classifier free guidance -
all conditions term. Defaults to 5.0.
cfg_coef_txt (float, optional): Coefficient used in multi-source classifier free guidance -
text condition term. Defaults to 0.0.
&#34;&#34;&#34;
self.generation_params = {
&#39;cfg_coef_all&#39;: cfg_coef_all,
&#39;cfg_coef_txt&#39;: cfg_coef_txt
}
self.generation_params.update(kwargs)
def _unnormalized_latents(self, latents: torch.Tensor) -&gt; torch.Tensor:
&#34;&#34;&#34;Unnormalize latents, shifting back to EnCodec&#39;s expected mean, std&#34;&#34;&#34;
assert self.cfg is not None
scaled = latents * self.cfg.compression_model_latent_std
return scaled + self.cfg.compression_model_latent_mean
def generate_audio(self, gen_latents: torch.Tensor) -&gt; torch.Tensor:
&#34;&#34;&#34;Decode audio from generated latents&#34;&#34;&#34;
assert gen_latents.dim() == 3 # [B, T, C]
# unnormalize latents
gen_latents = self._unnormalized_latents(gen_latents)
return self.compression_model.model.decoder(gen_latents.permute(0, 2, 1))
def _generate_tokens(self, attributes: tp.List[ConditioningAttributes],
prompt_tokens: tp.Optional[torch.Tensor], progress: bool = False) -&gt; torch.Tensor:
&#34;&#34;&#34;Generate continuous audio latents given conditions.
Args:
attributes (list of ConditioningAttributes): Conditions used for generation (here text).
prompt_tokens (torch.Tensor, optional): Audio prompt used for continuation.
progress (bool, optional): Flag to display progress of the generation process. Defaults to False.
Returns:
torch.Tensor: Generated latents, of shape [B, T, C].
&#34;&#34;&#34;
total_gen_len = int(self.duration * self.frame_rate)
max_prompt_len = int(min(self.duration, self.max_duration) * self.frame_rate)
def _progress_callback(ode_steps: int, max_ode_steps: int):
ode_steps += 1
if self._progress_callback is not None:
# Note that total_gen_len might be quite wrong depending on the
# codebook pattern used, but with delay it is almost accurate.
self._progress_callback(ode_steps, max_ode_steps)
else:
print(f&#39;{ode_steps: 6d} / {max_ode_steps: 6d}&#39;, end=&#39;\r&#39;)
if prompt_tokens is not None:
assert max_prompt_len &gt;= prompt_tokens.shape[-1], \
&#34;Prompt is longer than audio to generate&#34;
callback = None
if progress:
callback = _progress_callback
# generate by sampling from the LM
with self.autocast:
total_gen_len = math.ceil(self.duration * self.compression_model.frame_rate)
return self.lm.generate(
prompt_tokens, attributes,
callback=callback, max_gen_len=total_gen_len, **self.generation_params)
def _prepare_chord_conditions(
self,
attributes: tp.List[ConditioningAttributes],
chords: tp.Optional[tp.List[tp.Tuple[str, float]]],
) -&gt; tp.List[ConditioningAttributes]:
&#34;&#34;&#34;
Prepares chord conditions by translating symbolic chord progressions into a sequence of integers.
This method updates the ConditioningAttributes with per-frame chords information.
Args:
attributes (List[ConditioningAttributes]):
The initial attributes and optional tensor data.
chords (List[Tuple[str, float]]):
A list of tuples containing chord labels and their start times.
Returns:
List[ConditioningAttributes]:
The updated attributes with frame chords integrated, alongside the original optional tensor data.
&#34;&#34;&#34;
if chords is None or chords == []:
for att in attributes:
att.symbolic[JascoCondConst.CRD.value] = SymbolicCondition(frame_chords=-1 *
torch.ones(1, dtype=torch.int32))
return attributes
# flip from (chord, start_time) to (start_time, chord)
chords_time_first: tp.List[tuple[float, str]] = [(item[1], item[0]) for item in chords]
# translate symbolic chord progression into a sequence of ints
frame_chords = construct_frame_chords(min_timestamp=0,
chord_changes=chords_time_first,
mapping_dict=self.chords_mapping,
prev_chord=&#39;&#39;,
frame_rate=self.compression_model.frame_rate,
segment_duration=self.duration)
# update the attribute objects
for att in attributes:
att.symbolic[JascoCondConst.CRD.value] = SymbolicCondition(frame_chords=torch.tensor(frame_chords))
return attributes
@torch.no_grad()
def _prepare_drums_conditions(self,
attributes:
tp.List[ConditioningAttributes],
drums_wav: tp.Optional[torch.Tensor],
):
# prepare drums cond
for attr in attributes:
if drums_wav is None:
attr.wav[JascoCondConst.DRM.value] = WavCondition(
torch.zeros((1, 1, 1), device=self.device),
torch.tensor([0], device=self.device),
sample_rate=[self.sample_rate],
path=[None])
else:
if JascoCondConst.DRM.value not in self.lm.condition_provider.conditioners:
raise RuntimeError(&#34;This model doesn&#39;t support drums conditioning. &#34;)
expected_length = self.lm.cfg.dataset.segment_duration * self.sample_rate
# trim if needed
drums_wav = drums_wav[..., :expected_length]
# pad if needed
if drums_wav.shape[-1] &lt; expected_length:
diff = expected_length - drums_wav.shape[-1]
diff_zeros = torch.zeros((drums_wav.shape[0], drums_wav.shape[1], diff),
device=drums_wav.device, dtype=drums_wav.dtype)
drums_wav = torch.cat((drums_wav, diff_zeros), dim=-1)
attr.wav[JascoCondConst.DRM.value] = WavCondition(
drums_wav.to(device=self.device),
torch.tensor([drums_wav.shape[-1]], device=self.device),
sample_rate=[self.sample_rate],
path=[None],
)
return attributes
@torch.no_grad()
def _prepare_melody_conditions(
self,
attributes: tp.List[ConditioningAttributes],
melody: tp.Optional[torch.Tensor],
expected_length: int,
melody_bins: int = 53,
) -&gt; tp.List[ConditioningAttributes]:
&#34;&#34;&#34;
Prepares melody conditions by subtituting with pre-computed salience matrix.
This method updates the ConditioningAttributes with per-frame chords information.
Args:
attributes (List[ConditioningAttributes]):
The initial attributes and optional tensor data.
chords (List[Tuple[str, float]]):
A list of tuples containing chord labels and their start times.
Returns:
List[ConditioningAttributes]:
The updated attributes with frame chords integrated, alongside the original optional tensor data.
&#34;&#34;&#34;
for attr in attributes:
if melody is None:
melody = torch.zeros((melody_bins, expected_length))
attr.symbolic[JascoCondConst.MLD.value] = SymbolicCondition(melody=melody)
return attributes
@torch.no_grad()
def _prepare_temporal_conditions(
self,
attributes: tp.List[ConditioningAttributes],
expected_length: int,
chords: tp.Optional[tp.List[tp.Tuple[str, float]]],
drums_wav: tp.Optional[torch.Tensor],
salience_matrix: tp.Optional[torch.Tensor],
melody_bins: int = 53,
) -&gt; tp.List[ConditioningAttributes]:
&#34;&#34;&#34;
Prepares temporal conditions (chords, drums).
Args:
attributes (List[ConditioningAttributes]): The initial attributes and optional tensor data.
expected_length (int): The expected number of generated frames.
chords (List[Tuple[str, float]]): A list of tuples containing chord labels and their start times.
drums_wav (List[Tuple[str, float]]): tensor of extracted drums wav.
salience_matrix (List[Tuple[str, float]]): melody matrix.
melody_bins (int): number of melody bins the model was trained with, only relevant if trained with melody.
Returns:
List[ConditioningAttributes]:
The updated attributes after processing chord conditions.
&#34;&#34;&#34;
attributes = self._prepare_chord_conditions(attributes=attributes, chords=chords)
attributes = self._prepare_drums_conditions(attributes=attributes, drums_wav=drums_wav)
attributes = self._prepare_melody_conditions(attributes=attributes, melody=salience_matrix,
expected_length=expected_length, melody_bins=melody_bins)
return attributes
@torch.no_grad()
def generate_music(
self, descriptions: tp.List[str],
drums_wav: tp.Optional[torch.Tensor] = None,
drums_sample_rate: int = 32000,
chords: tp.Optional[tp.List[tp.Tuple[str, float]]] = None,
melody_salience_matrix: tp.Optional[torch.Tensor] = None,
iopaint_wav: tp.Optional[torch.Tensor] = None,
segment_duration: float = 10.0,
frame_rate: float = 50.0,
melody_bins: int = 53,
progress: bool = False, return_latents: bool = False) \
-&gt; tp.Union[torch.Tensor, tp.Tuple[torch.Tensor, torch.Tensor]]:
&#34;&#34;&#34;Generate samples conditioned on text and temporal conditions (chords, melody, drums).
Args:
descriptions (list of str): A list of strings used as text conditioning.
chords (list of (str, float) tuples): Chord progression represented as chord, start time (sec), e.g.:
[(&#34;C&#34;, 0.0), (&#34;F&#34;, 4.0), (&#34;G&#34;, 6.0), (&#34;C&#34;, 8.0)]
melody_salience_matrix (torch.Tensor, optional): melody saliency matrix. Default=None.
iopaint_wav (torch.Tensor, optional): in/out=painting waveform. Default=None.
segment_duration (float): the segment duration the model was trained on. Default=None.
frame_rate (float): the frame_rate model was trained on. Default=None.
melody_bins (int): number of melody bins the model was trained with, only relevant if trained with melody.
progress (bool, optional): Flag to display progress of the generation process. Defaults to False.
&#34;&#34;&#34;
if drums_wav is not None:
if drums_wav.dim() == 2:
drums_wav = drums_wav[None]
assert drums_wav.dim() == 3, &#34;drums wav should have a shape [B, C, T].&#34;
drums_wav = convert_audio(drums_wav, drums_sample_rate, self.sample_rate, self.audio_channels)
cond_attributes, prompt_tokens = self._prepare_tokens_and_attributes(descriptions=descriptions,
prompt=None)
# prepare temporal conds (symbolic / audio)
jasco_attributes = self._prepare_temporal_conditions(attributes=cond_attributes,
expected_length=int(segment_duration * frame_rate),
chords=chords,
drums_wav=drums_wav,
salience_matrix=melody_salience_matrix,
melody_bins=melody_bins)
assert prompt_tokens is None
tokens = self._generate_tokens(jasco_attributes, prompt_tokens, progress)
if return_latents:
return self.generate_audio(tokens), tokens
return self.generate_audio(tokens)
@torch.no_grad()
def generate(self, descriptions: tp.List[str], progress: bool = False, return_latents: bool = False) \
-&gt; tp.Union[torch.Tensor, tp.Tuple[torch.Tensor, torch.Tensor]]:
&#34;&#34;&#34;Generate samples conditioned on text.
Args:
descriptions (list of str): A list of strings used as text conditioning.
progress (bool, optional): Flag to display progress of the generation process. Defaults to False.
&#34;&#34;&#34;
return self.generate_music(descriptions=descriptions, progress=progress, return_latents=return_latents)</code></pre>
</details>
<div class="desc"><p>JASCO main model with convenient generation API.</p>
<h2 id="args">Args</h2>
<dl>
<dt><strong><code>chords_mapping_path</code></strong></dt>
<dd>path to chords to index mapping pickle</dd>
</dl>
<p>kwargs - See MusicGen class.</p></div>
<h3>Ancestors</h3>
<ul class="hlist">
<li><a title="audiocraft.models.genmodel.BaseGenModel" href="genmodel.html#audiocraft.models.genmodel.BaseGenModel">BaseGenModel</a></li>
<li>abc.ABC</li>
</ul>
<h3>Static methods</h3>
<dl>
<dt id="audiocraft.models.jasco.JASCO.get_pretrained"><code class="name flex">
<span>def <span class="ident">get_pretrained</span></span>(<span>name: str = 'facebook/jasco-chords-drums-400M',<br>device=None,<br>chords_mapping_path='assets/chord_to_index_mapping.pkl')</span>
</code></dt>
<dd>
<details class="source">
<summary>
<span>Expand source code</span>
</summary>
<pre><code class="python">@staticmethod
def get_pretrained(name: str = &#39;facebook/jasco-chords-drums-400M&#39;, device=None,
chords_mapping_path=&#39;assets/chord_to_index_mapping.pkl&#39;):
&#34;&#34;&#34;Return pretrained model, we provide 2 models:
1. facebook/jasco-chords-drums-400M: 10s music generation conditioned on
text, chords and drums, 400M parameters.
2. facebook/jasco-chords-drums-1B: 10s music generation conditioned on
text, chords and drums, 1B parameters.
&#34;&#34;&#34;
if device is None:
if torch.cuda.device_count():
device = &#39;cuda&#39;
else:
device = &#39;cpu&#39;
compression_model = load_compression_model(name, device=device)
lm = load_jasco_model(name, compression_model, device=device)
kwargs = {&#39;name&#39;: name,
&#39;compression_model&#39;: compression_model,
&#39;lm&#39;: lm,
&#39;chords_mapping_path&#39;: chords_mapping_path}
return JASCO(**kwargs)</code></pre>
</details>
<div class="desc"><p>Return pretrained model, we provide 2 models:
1. facebook/jasco-chords-drums-400M: 10s music generation conditioned on
text, chords and drums, 400M parameters.
2. facebook/jasco-chords-drums-1B: 10s music generation conditioned on
text, chords and drums, 1B parameters.</p></div>
</dd>
</dl>
<h3>Methods</h3>
<dl>
<dt id="audiocraft.models.jasco.JASCO.generate_audio"><code class="name flex">
<span>def <span class="ident">generate_audio</span></span>(<span>self, gen_latents: torch.Tensor) ‑> torch.Tensor</span>
</code></dt>
<dd>
<details class="source">
<summary>
<span>Expand source code</span>
</summary>
<pre><code class="python">def generate_audio(self, gen_latents: torch.Tensor) -&gt; torch.Tensor:
&#34;&#34;&#34;Decode audio from generated latents&#34;&#34;&#34;
assert gen_latents.dim() == 3 # [B, T, C]
# unnormalize latents
gen_latents = self._unnormalized_latents(gen_latents)
return self.compression_model.model.decoder(gen_latents.permute(0, 2, 1))</code></pre>
</details>
<div class="desc"><p>Decode audio from generated latents</p></div>
</dd>
<dt id="audiocraft.models.jasco.JASCO.generate_music"><code class="name flex">
<span>def <span class="ident">generate_music</span></span>(<span>self,<br>descriptions: List[str],<br>drums_wav: torch.Tensor | None = None,<br>drums_sample_rate: int = 32000,<br>chords: List[Tuple[str, float]] | None = None,<br>melody_salience_matrix: torch.Tensor | None = None,<br>iopaint_wav: torch.Tensor | None = None,<br>segment_duration: float = 10.0,<br>frame_rate: float = 50.0,<br>melody_bins: int = 53,<br>progress: bool = False,<br>return_latents: bool = False) ‑> torch.Tensor | Tuple[torch.Tensor, torch.Tensor]</span>
</code></dt>
<dd>
<details class="source">
<summary>
<span>Expand source code</span>
</summary>
<pre><code class="python">@torch.no_grad()
def generate_music(
self, descriptions: tp.List[str],
drums_wav: tp.Optional[torch.Tensor] = None,
drums_sample_rate: int = 32000,
chords: tp.Optional[tp.List[tp.Tuple[str, float]]] = None,
melody_salience_matrix: tp.Optional[torch.Tensor] = None,
iopaint_wav: tp.Optional[torch.Tensor] = None,
segment_duration: float = 10.0,
frame_rate: float = 50.0,
melody_bins: int = 53,
progress: bool = False, return_latents: bool = False) \
-&gt; tp.Union[torch.Tensor, tp.Tuple[torch.Tensor, torch.Tensor]]:
&#34;&#34;&#34;Generate samples conditioned on text and temporal conditions (chords, melody, drums).
Args:
descriptions (list of str): A list of strings used as text conditioning.
chords (list of (str, float) tuples): Chord progression represented as chord, start time (sec), e.g.:
[(&#34;C&#34;, 0.0), (&#34;F&#34;, 4.0), (&#34;G&#34;, 6.0), (&#34;C&#34;, 8.0)]
melody_salience_matrix (torch.Tensor, optional): melody saliency matrix. Default=None.
iopaint_wav (torch.Tensor, optional): in/out=painting waveform. Default=None.
segment_duration (float): the segment duration the model was trained on. Default=None.
frame_rate (float): the frame_rate model was trained on. Default=None.
melody_bins (int): number of melody bins the model was trained with, only relevant if trained with melody.
progress (bool, optional): Flag to display progress of the generation process. Defaults to False.
&#34;&#34;&#34;
if drums_wav is not None:
if drums_wav.dim() == 2:
drums_wav = drums_wav[None]
assert drums_wav.dim() == 3, &#34;drums wav should have a shape [B, C, T].&#34;
drums_wav = convert_audio(drums_wav, drums_sample_rate, self.sample_rate, self.audio_channels)
cond_attributes, prompt_tokens = self._prepare_tokens_and_attributes(descriptions=descriptions,
prompt=None)
# prepare temporal conds (symbolic / audio)
jasco_attributes = self._prepare_temporal_conditions(attributes=cond_attributes,
expected_length=int(segment_duration * frame_rate),
chords=chords,
drums_wav=drums_wav,
salience_matrix=melody_salience_matrix,
melody_bins=melody_bins)
assert prompt_tokens is None
tokens = self._generate_tokens(jasco_attributes, prompt_tokens, progress)
if return_latents:
return self.generate_audio(tokens), tokens
return self.generate_audio(tokens)</code></pre>
</details>
<div class="desc"><p>Generate samples conditioned on text and temporal conditions (chords, melody, drums).</p>
<h2 id="args">Args</h2>
<dl>
<dt><strong><code>descriptions</code></strong> :&ensp;<code>list</code> of <code>str</code></dt>
<dd>A list of strings used as text conditioning.</dd>
<dt>chords (list of (str, float) tuples): Chord progression represented as chord, start time (sec), e.g.:</dt>
<dt>[("C", 0.0), ("F", 4.0), ("G", 6.0), ("C", 8.0)]</dt>
<dt><strong><code>melody_salience_matrix</code></strong> :&ensp;<code>torch.Tensor</code>, optional</dt>
<dd>melody saliency matrix. Default=None.</dd>
<dt><strong><code>iopaint_wav</code></strong> :&ensp;<code>torch.Tensor</code>, optional</dt>
<dd>in/out=painting waveform. Default=None.</dd>
<dt><strong><code>segment_duration</code></strong> :&ensp;<code>float</code></dt>
<dd>the segment duration the model was trained on. Default=None.</dd>
<dt><strong><code>frame_rate</code></strong> :&ensp;<code>float</code></dt>
<dd>the frame_rate model was trained on. Default=None.</dd>
<dt><strong><code>melody_bins</code></strong> :&ensp;<code>int</code></dt>
<dd>number of melody bins the model was trained with, only relevant if trained with melody.</dd>
<dt><strong><code>progress</code></strong> :&ensp;<code>bool</code>, optional</dt>
<dd>Flag to display progress of the generation process. Defaults to False.</dd>
</dl></div>
</dd>
<dt id="audiocraft.models.jasco.JASCO.set_generation_params"><code class="name flex">
<span>def <span class="ident">set_generation_params</span></span>(<span>self, cfg_coef_all: float = 5.0, cfg_coef_txt: float = 0.0, **kwargs)</span>
</code></dt>
<dd>
<details class="source">
<summary>
<span>Expand source code</span>
</summary>
<pre><code class="python">def set_generation_params(self,
cfg_coef_all: float = 5.0,
cfg_coef_txt: float = 0.0,
**kwargs):
&#34;&#34;&#34;Set the generation parameters for JASCO.
Args:
cfg_coef_all (float, optional): Coefficient used in multi-source classifier free guidance -
all conditions term. Defaults to 5.0.
cfg_coef_txt (float, optional): Coefficient used in multi-source classifier free guidance -
text condition term. Defaults to 0.0.
&#34;&#34;&#34;
self.generation_params = {
&#39;cfg_coef_all&#39;: cfg_coef_all,
&#39;cfg_coef_txt&#39;: cfg_coef_txt
}
self.generation_params.update(kwargs)</code></pre>
</details>
<div class="desc"><p>Set the generation parameters for JASCO.</p>
<h2 id="args">Args</h2>
<dl>
<dt><strong><code>cfg_coef_all</code></strong> :&ensp;<code>float</code>, optional</dt>
<dd>Coefficient used in multi-source classifier free guidance -
all conditions term. Defaults to 5.0.</dd>
<dt><strong><code>cfg_coef_txt</code></strong> :&ensp;<code>float</code>, optional</dt>
<dd>Coefficient used in multi-source classifier free guidance -
text condition term. Defaults to 0.0.</dd>
</dl></div>
</dd>
</dl>
<h3>Inherited members</h3>
<ul class="hlist">
<li><code><b><a title="audiocraft.models.genmodel.BaseGenModel" href="genmodel.html#audiocraft.models.genmodel.BaseGenModel">BaseGenModel</a></b></code>:
<ul class="hlist">
<li><code><a title="audiocraft.models.genmodel.BaseGenModel.audio_channels" href="genmodel.html#audiocraft.models.genmodel.BaseGenModel.audio_channels">audio_channels</a></code></li>
<li><code><a title="audiocraft.models.genmodel.BaseGenModel.frame_rate" href="genmodel.html#audiocraft.models.genmodel.BaseGenModel.frame_rate">frame_rate</a></code></li>
<li><code><a title="audiocraft.models.genmodel.BaseGenModel.generate" href="genmodel.html#audiocraft.models.genmodel.BaseGenModel.generate">generate</a></code></li>
<li><code><a title="audiocraft.models.genmodel.BaseGenModel.generate_continuation" href="genmodel.html#audiocraft.models.genmodel.BaseGenModel.generate_continuation">generate_continuation</a></code></li>
<li><code><a title="audiocraft.models.genmodel.BaseGenModel.generate_unconditional" href="genmodel.html#audiocraft.models.genmodel.BaseGenModel.generate_unconditional">generate_unconditional</a></code></li>
<li><code><a title="audiocraft.models.genmodel.BaseGenModel.sample_rate" href="genmodel.html#audiocraft.models.genmodel.BaseGenModel.sample_rate">sample_rate</a></code></li>
<li><code><a title="audiocraft.models.genmodel.BaseGenModel.set_custom_progress_callback" href="genmodel.html#audiocraft.models.genmodel.BaseGenModel.set_custom_progress_callback">set_custom_progress_callback</a></code></li>
</ul>
</li>
</ul>
</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-classes">Classes</a></h3>
<ul>
<li>
<h4><code><a title="audiocraft.models.jasco.JASCO" href="#audiocraft.models.jasco.JASCO">JASCO</a></code></h4>
<ul class="">
<li><code><a title="audiocraft.models.jasco.JASCO.generate_audio" href="#audiocraft.models.jasco.JASCO.generate_audio">generate_audio</a></code></li>
<li><code><a title="audiocraft.models.jasco.JASCO.generate_music" href="#audiocraft.models.jasco.JASCO.generate_music">generate_music</a></code></li>
<li><code><a title="audiocraft.models.jasco.JASCO.get_pretrained" href="#audiocraft.models.jasco.JASCO.get_pretrained">get_pretrained</a></code></li>
<li><code><a title="audiocraft.models.jasco.JASCO.set_generation_params" href="#audiocraft.models.jasco.JASCO.set_generation_params">set_generation_params</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>
在新工单中引用 查看 Git Blame 复制永久链接