My Little Pony and My Hero Academia/僕のヒーローアカデミア crossover fanfiction archive with over 13 stories. Come in to read stories and fanfics that span multiple fandoms in the My Little Pony and My Hero Academia/僕のヒーローアカデミア universe.

后文用简称 MHA=Multi-Head Attention， PE=Positional Encoding。 Transfomer整体是Encoder-Decoder架构。 其中Encoder负责把源句子（如英文句子）编码为一种潜在的语义表示。 与 RNN/LSTM 不同，Transfomer中的编码是变长的，句子中每个词会有一个向量，整个句子的语义表示是一个向量的变长数组。 其中Decoder负责从对源句子的编码中解码出目标句子。 解码有两个步骤，一是对已经解码得到的目标词汇做编码，二是结合源句子编码和已经解码得到的上文， …

2019年6月30日 · Midoriya was a teenage boy in Musutafu, Japan. He lived in a world of superheroes… and supervillains. In his world over eighty percent of the people on Earth possessed some sort of superpower, a Quirk.

Izuku takes Bakugou's 'advice' and takes a swan dive off a roof. I got the idea from Flitty's story Six, which is a harry potter and mlp xover. A portion of the first chapter was ripped almost directly from their story.

2023年11月27日 · Transformer其实不是完全的Self-Attention（SA，自注意力）结构，还带有Cross-Attention（CA，交叉注意力）、残差连接、LayerNorm、类似1维卷积的Position-wise Feed-Forward Networks（FFN）、MLP和Positional Encoding（位置编码）等

2024年1月10日 · 本文介绍了一种简单直接的Transformer架构优化方法，以Transformer模型中的核心操作自注意力（SA）和交叉注意力层（CA）为优化目标，直接使用简单高效的MLP层进行替换。

A Quirk AU where Izuku gains the power to bring ponies into his world... by petting them. As his mom went out the front door, little 4 year old Izuku heard his mom said, "I'll only be gone to grocery store for 10 minutes. Don't open the door for any strangers." "Okay, Mom!" Izuku called out to her from the living room.

2022年4月4日 · In the world of Equestria there lived six pony Friends. And A Pegasus named rainbow dash was challenged to a race by none other than the Earth pony herself... pinkie pie! Within this Race, Rainbow Dash flew So fast that she Created a Triple Sonic Rainboom and traveled through too another Dimension !

Izuku Midoriya, aka Deku, finds himself in Camp Everfree and starts a series of adventures where he shows his heroic nature by being hero physically and emotionally. T My Little Academia: Friendship is Heroic! Twilight Sparkle wants to be a hero. She didn't expect the most powerful hero in Equestria to lend her a hoof.

2025年1月13日 · 为什么区分MHA和MLP？ 在Transformer模型中，Multi-Head Attention (MHA) 和 Multi-Layer Perceptron (MLP) 是两个核心组件，它们各自承担着不同的功能，共同协作以提升模型的性能。 MHA 的作用：