City
c52d82758a
Rename/move
2024-12-12 19:51:17 +01:00
City
3f844cac42
PixArt initial rewrite
2024-12-10 22:19:02 +01:00
a-One-Fan
8cff976421
Intel support for Pixart-Sigma ( #64 )
...
* Fix IPEX attn_drop
* IPEX SDPA 4GB fix
Alchemist GPUs lack the 64 bit emulation needed for >4GB in this case. Attention slicing workaround copied from https://github.com/vladmandic/automatic/blob/master/modules/intel/ipex/attention.py
* Do 4GB fix only when necessary
~20% performance boost to not use the 4GB fix when not necessary
* Move attention.py inside utils
* Better 4GB estimation
* Remove unnecessary warning
2024-07-06 20:26:29 +02:00
codeMODE
7d0cb8fe31
Merge branch 'city96:main' into get-depth-programatically
2024-06-23 19:43:25 +01:00
City
2ad0b09cb9
PixArt allow on-the-fly PE interpolation
2024-06-19 02:26:01 +02:00
gchapman
1d6c291e75
Minor optimization to stop a double .to() being needed
2024-06-18 21:45:10 +01:00
City
6867aacb26
PixArt fix float64 conversion error
...
#48
2024-05-25 18:09:56 +02:00
gchapman
0f888945f5
* Fix SDP attention mask by creating a diagonal mask in the right shape.
2024-05-16 09:23:16 +01:00
GavChap
3824b0d7b9
Update PixArt_blocks.py
...
Add dropout_p to SDP attention
2024-05-11 17:00:48 +01:00
City
c325dec0d3
PixArt sigma
...
Mentioned in #19
Also removes old example sampler since updating it would be too much overhead.
Pytorch attention is still wonky.
2024-04-07 17:55:33 +02:00
City
8befcdd9d6
PixArt controlnet test
2024-02-28 19:05:08 +01:00
City
80d5d9299b
Temp fix for mask shape
2023-12-08 17:23:51 +01:00
City
b31d8c517b
PixArt small fix
...
Issue #3 + some code changes from upstream.
2023-12-06 21:31:07 +01:00
City
5b42a090c4
xformers
2023-10-28 02:58:13 +02:00
City
814fe50852
PixArt support
2023-10-27 21:42:00 +02:00
City