Jump to content
Main menu
Main menu
move to sidebar
hide
Navigation
Main page
Recent changes
Random page
freem
Search
Search
Appearance
Create account
Log in
Personal tools
Create account
Log in
Pages for logged out editors
learn more
Contributions
Talk
Editing
训练音乐大模型
(section)
Add languages
Page
Discussion
English
Read
Edit
Edit source
View history
Tools
Tools
move to sidebar
hide
Actions
Read
Edit
Edit source
View history
General
What links here
Related changes
Special pages
Page information
Appearance
move to sidebar
hide
Warning:
You are not logged in. Your IP address will be publicly visible if you make any edits. If you
log in
or
create an account
, your edits will be attributed to your username, along with other benefits.
Anti-spam check. Do
not
fill this in!
=== Music Transformer === '''Music Transformer'''(2018年,Magenta团队)是早期将Transformer成功应用于音乐生成的里程碑项目 (Music Transformer: Generating Music with Long-Term Structure)。它针对音乐长结构的问题,引入'''相对位置编码'''的Transformer模型,可生成数分钟长且结构连贯的钢琴曲 (Music Transformer: Generating Music with Long-Term Structure) (Music Transformer: Generating Music with Long-Term Structure)。 * '''架构与特点''':基于Transformer解码器堆叠,自注意力机制能直接“看到”全曲已生成内容,不像LSTM那样受限于隐藏状态容量。这使其在捕捉乐曲重复、呼应等长程结构上表现突出 (Music Transformer: Generating Music with Long-Term Structure)。通过相对位置编码,模型学会关注音符之间的相对距离关系(如隔多少拍),从而能更好地延续节奏型和和声进行,即使生成长度超出训练段长度也能保持合理结构 (Music Transformer: Generating Music with Long-Term Structure)。 * '''数据与表示''':使用了MIDI数据(Magenta提供的Performance RNN数据,包含古典钢琴演奏),采用事件序列表示音符(音高、时值、力度等事件流) (Music Transformer: Generating Music with Long-Term Structure)。这种'''性能RNN风格'''的事件表示保留了演奏细节(力度、踩踏等),生成结果不仅有对的音符,还富有表现力。 * '''效果''':Music Transformer生成的钢琴曲被认为显著优于LSTM模型,在长时间和声连贯性方面尤其突出 (Music Transformer: Generating Music with Long-Term Structure)。演示曲目展示了模型能围绕某个动机发展出结构完整的音乐段落,而非杂乱音符拼凑。 * '''开源情况''':Magenta开源了该模型的TensorFlow实现和训练脚本,并提供了预训练模型权重,可以在Colab上试用 (Music Transformer: Generating Music with Long-Term Structure)。这对于想快速实验Transformer在音乐上的效果非常有帮助。我们可以借鉴其相对位置注意力和事件表示方案,在自己的模型中实现长程结构捕捉。 * '''局限''':Music Transformer主要生成'''钢琴独奏''',不支持多乐器。它输出MIDI,需要另行渲染成音频。而且受限于训练数据,风格主要是古典/浪漫派钢琴。对于现代编曲(鼓点、合成器)未作特殊优化。后续工作(如Pop Music Transformer)扩展了流行音乐的节奏表示 (Applications and Advances of Artificial Intelligence in Music Generation:A Review)。我们如需面向多乐器,可考虑在Music Transformer基础上增加轨道信息或改用多模态Transformer。 总的来说,Music Transformer证明了Transformer的强大建模能力,是'''符号音乐生成'''领域的重要baseline (Applications and Advances of Artificial Intelligence in Music Generation:A Review)。对于我们的项目,如果需要生成长结构的纯音乐序列,可以参考其架构。尤其相对注意力等技术细节,能够增强模型的长程记忆,这是处理音乐这种长序列数据的关键。
Summary:
Please note that all contributions to freem are considered to be released under the Creative Commons Attribution-ShareAlike 4.0 (see
Freem:Copyrights
for details). If you do not want your writing to be edited mercilessly and redistributed at will, then do not submit it here.
You are also promising us that you wrote this yourself, or copied it from a public domain or similar free resource.
Do not submit copyrighted work without permission!
Cancel
Editing help
(opens in new window)