Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Apr 20, 2023

• force einops 0.6.1 or greater and call allow_ops_in_compiled_graph

  lucidrains

  committed
  00e07b7

• remove einops exts for better pytorch 2.0 compile compatibility

  lucidrains

  committed
  0069857

Commits on Mar 7, 2023

• use .to(device) to avoid copy, within one_unet_in_gpu context

  lucidrains

  committed
  580274b

Commits on Mar 5, 2023

• always rederive the predicted noise from the clipped x0 for ddim + predict noise objective

  lucidrains

  committed
  848e8a4

Commits on Mar 4, 2023

• bump to newer package of clip-anytorch that allows for text encodings < maximum context length

  lucidrains

  committed
  cc58f75

Commits on Feb 12, 2023

• fix for self conditioning in diffusion prior network lucidrains#273

  lucidrains

  committed
  3b2cf7b

Commits on Dec 23, 2022

• default ddim sampling eta to 0

  lucidrains

  committed
  984d62a

Commits on Dec 15, 2022

• extra insurance in case eos id is not there

  lucidrains

  committed
  683dd98

Commits on Nov 23, 2022

• address lucidrains#266

  lucidrains

  committed
  067ac32

Commits on Nov 11, 2022

• bug fix cosine annealing optimizer in prior trainer (lucidrains#262)

  nousr

  authored
  91c8d1c

Commits on Nov 8, 2022

• depend on open-clip-torch (lucidrains#261)

  Show description for 08238a7
  nousr

  authored
  08238a7

Commits on Nov 7, 2022

• add open clip to train_config (lucidrains#260)

  Show description for 7166ad6
  nousr

  authored
  7166ad6

Commits on Oct 29, 2022

• bring in prediction of v objective, combining the findings from progressive distillation paper and imagen-video to the eventual extension of dalle2 to make-a-video

  lucidrains

  committed
  fbba0f9

Commits on Oct 25, 2022

• Add static graph param (lucidrains#226)

  Show description for 9f37705
  rom1504

  authored
  9f37705

Commits on Oct 23, 2022

• fix openclipadapter to be able to use latest open sourced sota model

  lucidrains

  committed
  c3df46e

Commits on Oct 19, 2022

• fix a dtype conversion issue for the diffusion timesteps in the diffusion prior, thanks to @JiaHeng-DLUT

  lucidrains

  committed
  41fabf2

Commits on Oct 18, 2022

• Fix assert message (lucidrains#253)

  JiaHeng-DLUT

  authored
  5975e82

Commits on Sep 29, 2022

• fix for use with larger openai clip models by extracting dimension of last layernorm in clip

  lucidrains

  committed
  c18c080

Commits on Sep 20, 2022

• fix readme dataloader example

  lucidrains

  authored
  b39653c

Commits on Sep 19, 2022

• show example of using SOTA open sourced open clip

  lucidrains

  committed
  39f8b6c

• handle open clip adapter image size being a tuple

  lucidrains

  committed
  d0c11b3

Commits on Sep 16, 2022

• Minor Decoder Train Script Fixes (lucidrains#242)

  Show description for 86e2d5b
  nousr

  authored
  86e2d5b

Commits on Sep 1, 2022

• in ddim, noise should be predicted after x0 is maybe clipped, thanks to @lukovnikov for pointing this out in another repository

  lucidrains

  committed
  0d82dff

• fix bug with misnamed variable in diffusion prior network

  lucidrains

  committed
  8bbc956

Commits on Aug 31, 2022

• todo

  lucidrains

  authored
  22019fd

• fix ddim to use alpha_cumprod

  lucidrains

  committed
  6fb7e91

Commits on Aug 28, 2022

• add two asserts to diffusion prior to ensure matching image embedding dimensions for clip, diffusion prior network, and what was set on diffusion prior

  lucidrains

  committed
  ba58ae0

Commits on Aug 25, 2022

• upgrade to best downsample

  lucidrains

  committed
  1cc5d0a

Commits on Aug 23, 2022

• fix classifier free guidance for diffusion prior, thanks to @jaykim9870 for spotting the issue

  lucidrains

  committed
  59fa101

Commits on Aug 20, 2022

• Merge pull request lucidrains#234 from Veldrovive/deepspeed_fp16

  Show description for 916ece1
  Veldrovive

  authored
  916ece1

• Fixed issues with clip and deepspeed fp16

  Show description for cbaadb6
  Veldrovive

  committed
  cbaadb6

• cast attention matrix back to original dtype pre-softmax in attention

  lucidrains

  committed
  083508f

Commits on Aug 19, 2022

• make it work for @ethancohen123

  lucidrains

  committed
  7762edd

Commits on Aug 17, 2022

• cite einops

  lucidrains

  authored
  de5e628

• gratitude

  lucidrains

  authored
  1b4046b

Pagination

PreviousNext