Commits · e1cb93c3835a4c1e975914b5ce898abf509b21fe · chenpangpang / ComfyUI

23 Feb, 2024 2 commits
- Fix model and cond transformer options merge · e1cb93c3
  logtd authored Feb 23, 2024
  
  e1cb93c3
- Cleanup uni_pc inpainting. · 10847dfa
  comfyanonymous authored Feb 23, 2024
```
This causes some small changes to the uni pc inpainting behavior but it
seems to improve results slightly.
```
  10847dfa
20 Feb, 2024 3 commits
- Add some latent2rgb matrices for previews. · 18c151b3
  comfyanonymous authored Feb 20, 2024
  
  18c151b3
- Pass pooled CLIP to stage b. · 0d0fbabd
  comfyanonymous authored Feb 20, 2024
  
  0d0fbabd
- Align simple scheduling closer to official stable cascade scheduler. · c6b7a157
  comfyanonymous authored Feb 20, 2024
  
  c6b7a157
19 Feb, 2024 6 commits
- Enable fp16 by default on mps. · 88f30040
  comfyanonymous authored Feb 19, 2024
  
  88f30040
- Remove print. · e93cdd0a
  comfyanonymous authored Feb 19, 2024
  
  e93cdd0a
- Support Stable Cascade in checkpoint format. · 3711b31d
  comfyanonymous authored Feb 19, 2024
  
  3711b31d
- Some cleanups to how the text encoders are loaded. · d91f45ef
  comfyanonymous authored Feb 19, 2024
  
  d91f45ef
- Forgot to commit this. · a7b5eaa7
  comfyanonymous authored Feb 19, 2024
  
  a7b5eaa7
- Support loading the Stable Cascade effnet and previewer as a VAE. · 3b2e5799
  comfyanonymous authored Feb 19, 2024
```
The effnet can be used to encode images for img2img with Stage C.
```
  3b2e5799
18 Feb, 2024 2 commits
- Fix gligen lowvram mode. · dccca1da
  comfyanonymous authored Feb 18, 2024
  
  dccca1da
- Add ModelSamplingStableCascade to control the shift sampling parameter. · 8b60d33b
  comfyanonymous authored Feb 18, 2024
```
shift is 2.0 by default on Stage C and 1.0 by default on Stage B.
```
  8b60d33b
17 Feb, 2024 10 commits
- Fix attention masks properly for multiple batches. · 6bcf57ff
  comfyanonymous authored Feb 17, 2024
  
  6bcf57ff
- fp8 weight support for Stable Cascade. · 11e3221f
  comfyanonymous authored Feb 17, 2024
  
  11e3221f
- Fix attention mask batch size in some attention functions. · f8706546
  comfyanonymous authored Feb 17, 2024
  
  f8706546
- Properly fix attention masks in CLIP with batches. · 3b9969c1
  comfyanonymous authored Feb 17, 2024
  
  3b9969c1
- Implement shift schedule for cascade stage C. · 5b40e7a5
  comfyanonymous authored Feb 17, 2024
  
  5b40e7a5
- Manual cast for bf16 on older GPUs. · 929e266f
  comfyanonymous authored Feb 17, 2024
  
  929e266f
- Fix clip attention mask issues on some hardware. · 6c875d84
  comfyanonymous authored Feb 17, 2024
  
  6c875d84
- Make Stable Cascade work on old pytorch 2.0 · 805c36ac
  comfyanonymous authored Feb 17, 2024
  
  805c36ac
- Support Stable Cascade Stage B lite. · f2d1d16f
  comfyanonymous authored Feb 16, 2024
  
  f2d1d16f
- Make --force-fp32 disable loading models in bf16. · 0b3c5048
  comfyanonymous authored Feb 16, 2024
  
  0b3c5048
16 Feb, 2024 5 commits
- StableCascade CLIP model support. · 97d03ae0
  comfyanonymous authored Feb 16, 2024
  
  97d03ae0
- Stable Cascade Stage B. · 667c9281
  comfyanonymous authored Feb 16, 2024
  
  667c9281
- Stable Cascade Stage C. · f83109f0
  comfyanonymous authored Feb 16, 2024
  
  f83109f0
- Stable Cascade Stage A. · 5e06baf1
  comfyanonymous authored Feb 16, 2024
  
  5e06baf1
- Small refactor of is_device_* functions. · aeaeca10
  comfyanonymous authored Feb 15, 2024
  
  aeaeca10
13 Feb, 2024 1 commit
- Don't init the CLIP model when the checkpoint has no CLIP weights. · 38b7ac6e
  comfyanonymous authored Feb 13, 2024
  
  38b7ac6e
09 Feb, 2024 2 commits
- Fix infinite while loop being possible in ddim_scheduler · f44225fd
  Jedrzej Kosinski authored Feb 09, 2024
  
  f44225fd
- Add a way to set different conditioning for the controlnet. · 25a4805e
  comfyanonymous authored Feb 09, 2024
  
  25a4805e
08 Feb, 2024 1 commit
- Allow custom samplers to request discard penultimate sigma · a352c021
  blepping authored Feb 08, 2024
  
  a352c021
07 Feb, 2024 1 commit
- Don't use numpy for calculating sigmas. · c661a8b1
  comfyanonymous authored Feb 07, 2024
  
  c661a8b1
05 Feb, 2024 2 commits
- Make minimum tile size the size of the overlap. · 236bda26
  comfyanonymous authored Feb 05, 2024
  
  236bda26
- Don't use is_bf16_supported to check for fp16 support. · 66e28ef4
  comfyanonymous authored Feb 04, 2024
  
  66e28ef4
04 Feb, 2024 1 commit
- Speed up SDXL on 16xx series with fp16 weights and manual cast. · 24129d78
  comfyanonymous authored Feb 04, 2024
  
  24129d78
02 Feb, 2024 1 commit
- Always use fp16 for the text encoders. · 4b023906
  comfyanonymous authored Feb 02, 2024
  
  4b023906
30 Jan, 2024 1 commit
- Put VAE key name in model config. · da7a8df0
  comfyanonymous authored Jan 30, 2024
  
  da7a8df0
26 Jan, 2024 1 commit
- Remove some unused imports. · 89507f8a
  comfyanonymous authored Jan 25, 2024
  
  89507f8a
23 Jan, 2024 1 commit
- typo fix - calculate_sigmas_scheduler (#2619) · 05cd0069
  Dr.Lt.Data authored Jan 23, 2024
```
self.scheduler -> scheduler_name
Co-authored-by: Lt.Dr.Data <lt.dr.data@gmail.com>
```
  05cd0069