wangbo/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2026-02-11 05:52:33 +08:00
Code Issues Actions 18 Packages Projects Releases Wiki Activity
1,389 Commits 95 Branches 118 Tags 1.3 GiB
e68beb56e4
Commit Graph

155 Commits

Author SHA1 Message Date
comfyanonymous
1e24a78d85 --gpu-only now keeps the VAE on the device. 2023-07-01 15:22:40 -04:00
comfyanonymous
2ee0aa317c Leave text_encoder on the CPU when it can handle it. 2023-07-01 14:38:51 -04:00
comfyanonymous
d5a7abe10d Try to keep text encoders loaded and patched to increase speed. 
load_model_gpu() is now used with the text encoder models instead of just
the unet.
 2023-07-01 13:28:07 -04:00
comfyanonymous
e946dca0e1 Make highvram and normalvram shift the text encoders to vram and back. 
This is faster on big text encoder models than running it on the CPU.
 2023-07-01 12:37:23 -04:00
comfyanonymous
7dff6c094c LoraLoader node now caches the lora file between executions. 2023-06-29 23:40:51 -04:00
comfyanonymous
790073a21d Move unet to device right after loading on highvram mode. 2023-06-29 20:43:06 -04:00
comfyanonymous
ae27a5625e Support for SDXL text encoder lora. 2023-06-28 02:22:49 -04:00
comfyanonymous
95008c22cd Add CheckpointSave node to save checkpoints. 
The created checkpoints contain workflow metadata that can be loaded by
dragging them on top of the UI or loading them with the "Load" button.

Checkpoints will be saved in fp16 or fp32 depending on the format ComfyUI
is using for inference on your hardware. To force fp32 use: --force-fp32

Anything that patches the model weights like merging or loras will be
saved.

The output directory is currently set to: output/checkpoints but that might
change in the future.
 2023-06-26 12:22:27 -04:00
comfyanonymous
0254a9d75b Support loras based on the stability unet implementation. 2023-06-26 02:56:11 -04:00
comfyanonymous
c4c2db6ead Add DualClipLoader to load clip models for SDXL. 
Update LoadClip to load clip models for SDXL refiner.
 2023-06-25 01:40:38 -04:00
comfyanonymous
818bae8e52 Fix CLIPLoader node. 2023-06-24 13:56:46 -04:00
comfyanonymous
0db33017af Add some more transformer hooks and move tomesd to comfy_extras. 
Tomesd now uses q instead of x to decide which tokens to merge because
it seems to give better results.
 2023-06-24 03:30:22 -04:00
comfyanonymous
a852d8b138 Move latent scale factor from VAE to model. 2023-06-23 02:33:31 -04:00
comfyanonymous
288f0c430d Fix bug when yaml config has no clip params. 2023-06-23 01:12:59 -04:00
comfyanonymous
40f218c4fa Fix error with ClipVision loader node. 2023-06-23 01:08:05 -04:00
comfyanonymous
4ed8aea1a1 Don't merge weights when shapes don't match and print a warning. 2023-06-22 19:08:31 -04:00
comfyanonymous
08f1f7686c Support base SDXL and SDXL refiner models. 
Large refactor of the model detection and loading code.
 2023-06-22 13:03:50 -04:00
comfyanonymous
cd8d0b73c5 Fix last commits causing an issue with the text encoder lora. 2023-06-20 19:44:39 -04:00
comfyanonymous
af9e05f389 Keep a set of model_keys for faster add_patches. 2023-06-20 19:08:48 -04:00
comfyanonymous
2c71c47ff9 Add a type of model patch useful for model merging. 2023-06-20 17:34:11 -04:00
comfyanonymous
873b08bd0f Add a way to set patches that modify the attn2 output. 
Change the transformer patches function format to be more future proof.
 2023-06-18 22:58:22 -04:00
comfyanonymous
78eabd0fd4 This is not needed anymore and causes issues with alphas_cumprod. 2023-06-18 03:18:25 -04:00
comfyanonymous
282638b813 Add a --gpu-only argument to keep and run everything on the GPU. 
Make the CLIP model work on the GPU.
 2023-06-15 15:38:52 -04:00
comfyanonymous
0a80429175 Set model to fp16 before loading the state dict to lower ram bump. 2023-06-14 12:48:02 -04:00
comfyanonymous
fd8fa51a6d sampler_cfg_function now uses a dict for the argument. 
This means arguments can be added without issues.
 2023-06-13 16:10:36 -04:00
comfyanonymous
e4e6a23fec Turn on safe load for a few models. 2023-06-13 10:12:03 -04:00
comfyanonymous
ce4e360edf Cleanup: Remove a bunch of useless files. 2023-06-13 02:19:08 -04:00
comfyanonymous
a2bc14b56f Split the batch in VAEEncode if there's not enough memory. 2023-06-12 00:21:50 -04:00
comfyanonymous
8ef4de36f4 Auto switch to tiled VAE encode if regular one runs out of memory. 2023-06-11 23:25:39 -04:00
comfyanonymous
1e0ad9564e Simpler base model code. 2023-06-09 12:31:16 -04:00
comfyanonymous
6c6ef17bd4 Small refactor. 2023-06-06 13:23:01 -04:00
comfyanonymous
d7821166b2 Implement global average pooling for controlnet. 2023-06-03 01:49:03 -04:00
comfyanonymous
fa734798f1 This is useless for inference. 2023-05-31 13:03:24 -04:00
comfyanonymous
4af4fe017b Auto load model in lowvram if not enough memory. 2023-05-30 12:36:41 -04:00
comfyanonymous
12e275ab18 Support VAEs in diffusers format. 2023-05-28 02:02:09 -04:00
BlenderNeko
3cbb3ef058 comment out annoying print statement 2023-05-12 23:57:40 +02:00
BlenderNeko
2e2c17131b minor changes for tiled sampler 2023-05-12 23:49:09 +02:00
comfyanonymous
2edaaba3c2 Fix imports. 2023-05-04 18:10:29 -04:00
comfyanonymous
10ff210ffb Refactor. 2023-05-03 17:48:35 -04:00
pythongosssss
eaeac55c0d remove unused import 2023-05-03 18:21:23 +01:00
pythongosssss
d8017626fb use comfy progress bar 2023-05-03 18:19:22 +01:00
pythongosssss
f6154607f9 Merge remote-tracking branch 'origin/master' into tiled-progress 2023-05-03 17:33:42 +01:00
pythongosssss
8160309db9 reduce duplication 2023-05-03 17:33:19 +01:00
pythongosssss
33b0ba6464 added progress to encode + upscale 2023-05-02 19:18:07 +01:00
comfyanonymous
55236dc812 LoKR support. 2023-05-01 18:18:23 -04:00
pythongosssss
51a069805b Add progress to vae decode tiled 2023-04-24 11:55:44 +01:00
comfyanonymous
e6771d0986 Implement Linear hypernetworks. 
Add a HypernetworkLoader node to use hypernetworks.
 2023-04-23 12:35:25 -04:00
comfyanonymous
6c156642e4 Add support for GLIGEN textbox model. 2023-04-19 11:06:32 -04:00
comfyanonymous
3fe1252a35 Add a way for nodes to set a custom CFG function. 2023-04-17 11:05:15 -04:00
comfyanonymous
4df70d0f62 Fix model_management import so it doesn't get executed twice. 2023-04-15 19:04:33 -04:00
comfyanonymous
f089d4abc7 Some refactoring: from_tokens -> encode_from_tokens 2023-04-15 18:46:58 -04:00
BlenderNeko
47b2d342a8 ensure backwards compat with optional args 2023-04-14 21:16:55 +02:00
BlenderNeko
928cb6b2c7 split tokenizer from encoder 2023-04-13 22:06:50 +02:00
comfyanonymous
b55667284c Add support for unCLIP SD2.x models. 
See _for_testing/unclip in the UI for the new nodes.

unCLIPCheckpointLoader is used to load them.

unCLIPConditioning is used to add the image cond and takes as input a
CLIPVisionEncode output which has been moved to the conditioning section.
 2023-04-01 23:19:15 -04:00
comfyanonymous
7da8d5f9f5 Add a TomePatchModel node to the _for_testing section. 
Tome increases sampling speed at the expense of quality.
 2023-03-31 17:19:58 -04:00
comfyanonymous
7e2f4b0897 Split VAE decode batches depending on free memory. 2023-03-29 02:24:37 -04:00
comfyanonymous
121beab586 Support loha that use cp decomposition. 2023-03-23 04:32:25 -04:00
comfyanonymous
d15c9f2c3b Add loha support. 2023-03-23 03:40:12 -04:00
comfyanonymous
7cae5f5769 Try again with vae tiled decoding if regular fails because of OOM. 2023-03-22 14:49:00 -04:00
comfyanonymous
d8d65e7c9f Less seams in tiled outputs at the cost of more processing. 2023-03-22 03:29:09 -04:00
comfyanonymous
921f4e14a4 Add support for locon mid weights. 2023-03-21 14:51:51 -04:00
comfyanonymous
29728e1119 Try to fix a vram issue with controlnets. 2023-03-19 10:50:38 -04:00
comfyanonymous
d5bf8038c8 Merge T2IAdapterLoader and ControlNetLoader. 
Workflows will be auto updated.
 2023-03-17 18:17:59 -04:00
comfyanonymous
889689c7b2 use half() on fp16 models loaded with config. 2023-03-13 21:12:48 -04:00
comfyanonymous
edba761868 Use half() function on model when loading in fp16. 2023-03-13 20:58:09 -04:00
comfyanonymous
5be28c4069 Remove omegaconf dependency and some ci changes. 2023-03-13 14:49:18 -04:00
comfyanonymous
d777b2f4a9 Add a VAEEncodeTiled node. 2023-03-11 15:28:15 -05:00
comfyanonymous
77290f7110 Tiled upscaling with the upscale models. 2023-03-11 14:04:13 -05:00
comfyanonymous
66cca3659d Add locon support. 2023-03-09 21:41:24 -05:00
comfyanonymous
fa4dda2550 SD2.x controlnets now work. 2023-03-08 01:13:38 -05:00
comfyanonymous
341ebaabd8 Relative imports to test something. 2023-03-07 11:00:35 -05:00
comfyanonymous
339b2533eb Fix clip_skip no longer being loaded from yaml file. 2023-03-06 11:34:02 -05:00
comfyanonymous
ca7e2e3827 Add --cpu to use the cpu for inference. 2023-03-06 10:50:50 -05:00
comfyanonymous
663fa6eafd Implement support for t2i style model. 
It needs the CLIPVision model so I added CLIPVisionLoader and CLIPVisionEncode.

Put the clip vision model in models/clip_vision
Put the t2i style model in models/style_models

StyleModelLoader to load it, StyleModelApply to apply it
ConditioningAppend to append the conditioning it outputs to a positive one.
 2023-03-05 18:39:25 -05:00
comfyanonymous
7ee5e2d40e Add support for new colour T2I adapter model. 2023-03-03 19:13:07 -05:00
comfyanonymous
c195dab61c Add a node to set CLIP skip. 
Use a more simple way to detect if the model is -v prediction.
 2023-03-03 13:04:36 -05:00
comfyanonymous
5608730809 To be really simple CheckpointLoaderSimple should pick the right type. 2023-03-03 11:07:10 -05:00
comfyanonymous
8a2699b47d New CheckpointLoaderSimple to load checkpoints without a config. 2023-03-03 03:37:35 -05:00
comfyanonymous
a59bb36cb8 Try to improve memory issues with del. 2023-02-28 12:27:43 -05:00
comfyanonymous
7add9fe7b3 Preparing to add another function to load checkpoints. 2023-02-26 17:29:01 -05:00
comfyanonymous
084f8084ed Fix multiple controlnets not working. 2023-02-25 22:12:22 -05:00
comfyanonymous
9f1ca0847b Fixed issue when batched image was used as a controlnet input. 2023-02-25 14:57:28 -05:00
comfyanonymous
7f31625158 Fix missing variable. 2023-02-25 12:19:03 -05:00
comfyanonymous
4a31343a3d Add a T2IAdapterLoader node to load T2I-Adapter models. 
They are loaded as CONTROL_NET objects because they are similar.
 2023-02-25 01:24:56 -05:00
comfyanonymous
5cb9c83936 Prepare for t2i adapter. 2023-02-24 23:36:17 -05:00
comfyanonymous
a4778b6a31 Added an experimental VAEDecodeTiled. 
This decodes the image with the VAE in tiles which should be faster and
use less vram.

It's in the _for_testing section so I might change/remove it or even
add the functionality to the regular VAEDecode node depending on how
well it performs which means don't depend too much on it.
 2023-02-24 02:10:10 -05:00
comfyanonymous
f1b122ba02 Add a node to load diff controlnets. 2023-02-22 23:22:03 -05:00
comfyanonymous
9cd0189e7e ControlNetApply now stacks. 
It can be used to apply multiple control nets at the same time.
 2023-02-21 01:18:53 -05:00
comfyanonymous
7dd65e43dd Low vram mode for controlnets. 2023-02-17 15:48:16 -05:00
comfyanonymous
7010bf9e60 Use fp16 for fp16 control nets. 2023-02-17 15:31:38 -05:00
comfyanonymous
9999b1e198 Add a way to control controlnet strength. 2023-02-16 18:08:01 -05:00
comfyanonymous
b5b68268ee Add ControlNet support. 2023-02-16 10:38:08 -05:00
comfyanonymous
a84cd0d1ad Don't unload/reload model from CPU uselessly. 2023-02-08 03:40:43 -05:00
comfyanonymous
b1a7c9ebf6 Embeddings/textual inversion support for SD2.x 2023-02-05 15:49:03 -05:00
comfyanonymous
1de5aa6a59 Add a CLIPLoader node to load standalone clip weights. 
Put them in models/clip
 2023-02-05 15:20:18 -05:00
comfyanonymous
56d802e1f3 Use transformers CLIP instead of open_clip for SD2.x 
This should make things a bit cleaner.
 2023-02-05 14:36:28 -05:00
comfyanonymous
bf9ccffb17 Small fix for SD2.x loras. 2023-02-05 11:38:25 -05:00
comfyanonymous
678105fade SD2.x CLIP support for Loras. 2023-02-05 01:54:09 -05:00
comfyanonymous
ef90e9c376 Add a LoraLoader node to apply loras to models and clip. 
The models are modified in place before being used and unpatched after.
I think this is better than monkeypatching since it might make it easier
to use faster non pytorch unet inference in the future.
 2023-02-03 02:46:24 -05:00
comfyanonymous
f73e57d881 Add support for textual inversion embedding for SD1.x CLIP. 2023-01-29 18:46:44 -05:00