6fbfd19f05
fix dev branch by reverting
2022-09-21 19:41:42 -07:00
c3560d8e8e
Add files via upload
2022-09-21 07:58:21 +02:00
7b62670137
Add files via upload
2022-09-21 07:57:52 +02:00
782bb2e824
Add files via upload
2022-09-21 07:57:35 +02:00
9cc4816c2e
Add files via upload
2022-09-21 07:56:34 +02:00
7710ae423f
Add files via upload
2022-09-21 07:56:03 +02:00
88bf28a0ed
Merge with Doggettx's stable-diffusion fork
...
Allows for running at higher resolutions without running out of VRAM.
2022-09-10 16:06:01 -05:00
c2d72c5c23
Missed one function to revert
2022-09-07 09:03:19 +02:00
830f6946f9
Reverted in place tensor functions back to CompVis version
...
Improves performance and is no longer needed.
2022-09-07 08:43:36 +02:00
f134e245ba
Added max. res info to memory exception
2022-09-06 10:43:19 +02:00
755bec8923
Raise error when steps too high
...
Technically you could run at higher steps as long as the resolution is dividable by the steps but you're going to run into memory issues later on anyhow.
2022-09-06 09:39:10 +02:00
7a32fd6493
Commented out debug info
...
Forgot to comment out debug info
2022-09-06 09:12:09 +02:00
507ddec578
Fixed free memory calculation
...
Old version gave incorrect free memory results causing in crashes on edge cases.
2022-09-06 09:08:46 +02:00
d3c91ec937
Fixed memory handling for model.decode_first_stage
...
Better memory handling for model.decode_first_stage so it doesn't crash anymore after 100% rendering
2022-09-05 19:49:45 +02:00
8283bb5b84
Update attention.py
2022-09-05 12:03:59 +02:00
5065b41ce1
Update attention.py
...
Correction to comment
2022-09-05 10:13:59 +02:00
afaefd2e5f
Update attention.py
...
Run attention in a loop to allow for much higher resolutions (over 1920x1920 on a 3090)
2022-09-05 09:26:27 +02:00
1c25d0c6b1
Updating ipynb with colab-convert. Changing ddim schedule to include the 0 sigma noise step.
2022-09-02 12:26:26 -05:00
04b56c84a1
Added a new ddim discretization method called fill that fixes out of bounds errors. Embracing k-diffusion sigmas for the mask schedule
2022-08-31 00:29:49 -05:00
fe8810319b
Fixed missmatching steps. Strength can be used without an init image.
2022-08-29 13:18:17 -05:00
3472233c58
Merge https://github.com/enzymezoo-code/stable-diffusion
2022-08-22 15:31:44 -05:00
6a39340bb0
The img_callback function in all the samplers should take in img, not pred_x0. This allows us to pass a function to the sampler that can mutate the image at each step, and is also more consistent with the img_callback function in ddpm.py
2022-08-21 21:16:03 -05:00
c48f365695
Update plms.py
2022-08-19 17:20:17 -07:00
0258d85766
Update ddpm.py
2022-08-19 17:19:53 -07:00
a2ed19663a
Update ddim.py
2022-08-19 17:19:23 -07:00
277920a133
Update plms.py
2022-08-17 23:17:03 -07:00
0d221e264f
Update ddpm.py
2022-08-17 23:16:46 -07:00
8953eca484
Update ddim.py
2022-08-17 23:16:26 -07:00
2ff270f4e0
stable diffusion
2022-08-10 16:30:49 +02:00
f13bf9bf46
add vqgan loss with codebook statistic eval
2022-02-21 15:06:50 +01:00
2b46bcb98c
Update ddpm.py
...
clean up no.1
2022-01-17 21:24:19 +01:00
171cf29fb5
add configs for training unconditional/class-conditional ldms
2021-12-22 15:57:23 +01:00
e66308c7f2
add code
2021-12-21 03:23:41 +01:00
deforum
nicolai256
pharmapsychotic
Doggettx
enzymezoo-code
deforum
Robin Rombach
Robin Rombach
ablattmann