Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Dec 27, 2024

• add hyper connections

  lucidrains

  committed
  17a3fd2

• address #4

  lucidrains

  committed
  8e63355

Commits on Aug 24, 2024

• for adaptive attention, allow each head to have different adaptive weight gating

  lucidrains

  committed
  2c7137e

Commits on Jul 6, 2024

• offer attention softclamping

  lucidrains

  committed
  53e0106

• address #1

  lucidrains

  committed
  47e0222

Commits on May 10, 2024

• credit assignment

  lucidrains

  committed
  c796ddd

• complete sidequest and bump to 0.1.0

  lucidrains

  committed
  57a3c59

Commits on May 7, 2024

• remove wip

  lucidrains

  authored
  34d6ea8

Commits on May 6, 2024

• handle the final norm at the end of the transformer, but make optional

  lucidrains

  committed
  8e20163

• ability to add a few register tokens

  lucidrains

  committed
  5f207e0

Commits on May 5, 2024

• fix logic for skipping last feedforward in mmdit

  lucidrains

  committed
  7885107

• make sure adaptive layernorm film is identity in generalized version

  lucidrains

  committed
  49ec2fe

• remove Optional

  lucidrains

  committed
  15fdfa4

• add a generalized version for N modalities

  lucidrains

  committed
  c201077

• save on the last text feedforward

  lucidrains

  committed
  119869e

• naming

  lucidrains

  committed
  0b0be99

Commits on May 4, 2024

• oops

  lucidrains

  committed
  d1f967b

• film should be initialized to identity

  lucidrains

  committed
  263b76b

• able to use flash attention

  lucidrains

  committed
  f3301e0

• complete mmdit block and transformer for pytorch first

  lucidrains

  committed
  ca0e17b

• intent to implement

  lucidrains

  authored
  6aee625

• diagram

  lucidrains

  committed
  de123f3

• cite

  lucidrains

  authored
  60b5acf

• Initial commit

  lucidrains

  authored
  a93fd14