Xiangxi Guo (Ryan)
f0d5d0111f
Avoid torch compile graphbreak for older pytorch versions ( #9344 )
...
Turns out torch.compile has some gaps in context manager decorator
syntax support. I've sent patches to fix that in PyTorch, but it won't
be available for all the folks running older versions of PyTorch, hence
this trivial patch.
2025-08-14 23:41:37 -04:00
comfyanonymous
9df8792d4b
Make last PR not crash comfy on old pytorch. ( #9324 )
2025-08-13 15:12:41 -04:00
contentis
3da5a07510
SDPA backend priority ( #9299 )
2025-08-13 14:53:27 -04:00
doctorpangloss
69a4906964
Experimental GGUF support
2025-07-28 17:02:20 -07:00
doctorpangloss
04e411c32e
Merge branch 'master' of github.com:comfyanonymous/ComfyUI
2025-07-14 13:45:09 -07:00
comfyanonymous
111f583e00
Fallback to regular op when fp8 op throws exception. ( #8761 )
2025-07-02 00:57:13 -04:00
doctorpangloss
82388d51a2
Merge branch 'master' of github.com:comfyanonymous/ComfyUI
2025-06-17 10:35:10 -07:00
comfyanonymous
d42613686f
Fix issue with fp8 ops on some models. ( #8045 )
...
_scaled_mm errors when an input is non contiguous.
2025-05-10 07:52:56 -04:00
comfyanonymous
ac10a0d69e
Make loras work with --async-offload ( #7824 )
2025-04-26 19:56:22 -04:00
comfyanonymous
0dcc75ca54
Add experimental --async-offload lowvram weight offloading. ( #7820 )
...
This should speed up the lowvram mode a bit. It currently is only enabled when --async-offload is used but it will be enabled by default in the future if there are no problems.
2025-04-26 16:11:21 -04:00
doctorpangloss
5823497d55
Merge branch 'master' of github.com:comfyanonymous/ComfyUI
2025-04-21 13:14:36 -07:00
comfyanonymous
9ad792f927
Basic support for hidream i1 model.
2025-04-15 17:35:05 -04:00
comfyanonymous
8a438115fb
add RMSNorm to comfy.ops
2025-04-14 18:00:33 -04:00
catboxanon
1714a4c158
Add CublasOps support ( #7574 )
...
* CublasOps support
* Guard CublasOps behind --fast arg
2025-04-12 18:29:15 -04:00
doctorpangloss
040a324346
Merge branch 'master' of github.com:comfyanonymous/ComfyUI
2025-03-29 15:57:24 -07:00
comfyanonymous
70e15fd743
No need for scale_input when fp8 matrix mult is disabled.
2025-03-07 04:49:20 -05:00
comfyanonymous
e1474150de
Support fp8_scaled diffusion models that don't use fp8 matrix mult.
2025-03-07 04:39:21 -05:00
doctorpangloss
3c82be86d1
Merge branch 'master' of github.com:comfyanonymous/ComfyUI
2025-03-05 14:38:50 -08:00
comfyanonymous
4dc6709307
Rename argument in last commit and document the options.
2025-03-01 02:43:49 -05:00
Chenlei Hu
4d55f16ae8
Use enum list for --fast options ( #7024 )
2025-03-01 02:37:35 -05:00
comfyanonymous
cf0b549d48
--fast now takes a number as argument to indicate how fast you want it.
...
The idea is that you can indicate how much quality vs speed you want.
At the moment:
--fast 2 enables fp16 accumulation if your pytorch supports it.
--fast 5 enables fp8 matrix mult on fp8 models and the optimization above.
--fast without a number enables all optimizations.
2025-02-28 02:48:20 -05:00
doctorpangloss
693038738a
Merge branch 'master' of github.com:comfyanonymous/ComfyUI
2025-02-24 09:39:26 -08:00
comfyanonymous
ab888e1e0b
Add add_weight_wrapper function to model patcher.
...
Functions can now easily be added to wrap/modify model weights.
2025-02-12 05:55:35 -05:00
doctorpangloss
9d5a5dd533
Merge branch 'master' of github.com:comfyanonymous/ComfyUI
2024-12-28 14:24:27 -08:00
comfyanonymous
99a1fb6027
Make fast fp8 take a bit less peak memory.
2024-12-24 18:05:19 -05:00
doctorpangloss
2d1676c717
Merge branch 'master' of github.com:comfyanonymous/ComfyUI
2024-12-09 15:54:37 -08:00
Haoming
fbf68c4e52
clamp input ( #5928 )
2024-12-07 14:00:31 -05:00
doctorpangloss
31eacb6ac9
Improve compilation of models, adding support for triton
2024-11-01 10:40:58 -07:00
doctorpangloss
a8d8bff548
Improve support for torch compilation and sage attention
2024-10-29 19:22:26 -07:00
doctorpangloss
76a80a65ea
Merge branch 'master' of github.com:comfyanonymous/ComfyUI
2024-10-29 15:35:39 -07:00
comfyanonymous
915fdb5745
Fix lowvram edge case.
2024-10-22 16:34:50 -04:00
comfyanonymous
8ce2a1052c
Optimizations to --fast and scaled fp8.
2024-10-22 02:12:28 -04:00
comfyanonymous
0075c6d096
Mixed precision diffusion models with scaled fp8.
...
This change allows supports for diffusion models where all the linears are
scaled fp8 while the other weights are the original precision.
2024-10-21 18:12:51 -04:00
comfyanonymous
83ca891118
Support scaled fp8 t5xxl model.
2024-10-20 22:27:00 -04:00
comfyanonymous
f9f9faface
Fixed model merging issue with scaled fp8.
2024-10-20 06:24:31 -04:00
comfyanonymous
a68bbafddb
Support diffusion models with scaled fp8 weights.
2024-10-19 23:47:42 -04:00
comfyanonymous
67158994a4
Use the lowvram cast_to function for everything.
2024-10-17 17:25:56 -04:00
doctorpangloss
8512f361fe
Merge branch 'master' of github.com:comfyanonymous/ComfyUI
2024-10-14 15:26:27 -07:00
doctorpangloss
f3da381869
Fix inference mode execution issues
2024-10-10 21:00:09 -07:00
doctorpangloss
a38968f098
Improvements to execution
...
- Validation errors that occur early in the lifecycle of prompt
execution now get propagated to their callers in the
EmbeddedComfyClient. This includes error messages about missing node
classes.
- The execution context now includes the node_id and the prompt_id
- Latent previews are now sent with a node_id. This is not backwards
compatible with old frontends.
- Dependency execution errors are now modeled correctly.
- Distributed progress encodes image previews with node and prompt IDs.
- Typing for models
- The frontend was updated to use node IDs with previews
- Improvements to torch.compile experiments
- Some controlnet_aux nodes were upstreamed
2024-10-10 19:30:18 -07:00
doctorpangloss
69e523b89d
Experimental quantization support. Only Linux is meaningfully supported
2024-10-10 13:43:06 -07:00
comfyanonymous
e38c94228b
Add a weight_dtype fp8_e4m3fn_fast to the Diffusion Model Loader node.
...
This is used to load weights in fp8 and use fp8 matrix multiplication.
2024-10-09 19:43:17 -04:00
doctorpangloss
fa3176f96f
Merge branch 'master' of github.com:comfyanonymous/ComfyUI
2024-09-23 12:50:31 -07:00
comfyanonymous
9c41bc8d10
Remove useless line.
2024-09-23 02:32:29 -04:00
comfyanonymous
dc96a1ae19
Load controlnet in fp8 if weights are in fp8.
2024-09-21 04:50:12 -04:00
doctorpangloss
5155a3e248
Merge WIP
2024-08-25 18:52:29 -07:00
comfyanonymous
8ae23d8e80
Fix onnx export.
2024-08-23 17:52:47 -04:00
comfyanonymous
c7ee4b37a1
Try to fix some lora issues.
2024-08-22 15:32:18 -04:00
comfyanonymous
904bf58e7d
Make --fast work on pytorch nightly.
2024-08-21 14:01:41 -04:00
Svein Ove Aas
5f50263088
Replace use of .view with .reshape ( #4522 )
...
When generating images with fp8_e4_m3 Flux and batch size >1, using --fast, ComfyUI throws a "view size is not compatible with input tensor's size and stride" error pointing at the first of these two calls to view.
As reshape is semantically equivalent to view except for working on a broader set of inputs, there should be no downside to changing this. The only difference is that it clones the underlying data in cases where .view would error out. I have confirmed that the output still looks as expected, but cannot confirm that no mutable use is made of the tensors anywhere.
Note that --fast is only marginally faster than the default.
2024-08-21 11:21:48 -04:00
comfyanonymous
03ec517afb
Remove useless line, adjust windows default reserved vram.
2024-08-21 00:47:19 -04:00
comfyanonymous
510f3438c1
Speed up fp8 matrix mult by using better code.
2024-08-20 22:53:26 -04:00
comfyanonymous
9953f22fce
Add --fast argument to enable experimental optimizations.
...
Optimizations that might break things/lower quality will be put behind
this flag first and might be enabled by default in the future.
Currently the only optimization is float8_e4m3fn matrix multiplication on
4000/ADA series Nvidia cards or later. If you have one of these cards you
will see a speed boost when using fp8_e4m3fn flux for example.
2024-08-20 11:55:51 -04:00
comfyanonymous
538cb068bc
Make cast_to a nop if weight is already good.
2024-08-20 10:46:36 -04:00
comfyanonymous
39f114c44b
Less broken non blocking?
2024-08-18 16:53:17 -04:00
comfyanonymous
6730f3e1a3
Disable non blocking.
...
It fixed some perf issues but caused other issues that need to be debugged.
2024-08-18 14:38:09 -04:00
comfyanonymous
73332160c8
Enable non blocking transfers in lowvram mode.
2024-08-18 10:29:33 -04:00
doctorpangloss
0a1ae64b0b
Merge branch 'master' of github.com:comfyanonymous/ComfyUI
2024-08-01 16:19:11 -07:00
comfyanonymous
b85216a3c0
Lower T5 memory usage by a few hundred MB.
2024-07-31 00:52:34 -04:00
comfyanonymous
25853d0be8
Use common function for casting weights to input.
2024-07-30 10:49:14 -04:00
doctorpangloss
8cdc246450
Merge branch 'master' of github.com:comfyanonymous/ComfyUI
2024-06-17 16:19:48 -07:00
comfyanonymous
bb1969cab7
Initial support for the stable audio open model.
2024-06-15 12:14:56 -04:00
doctorpangloss
cb557c960b
Merge branch 'master' of github.com:comfyanonymous/ComfyUI
2024-05-31 07:42:11 -07:00
comfyanonymous
6c23854f54
Fix OSX latent2rgb previews.
2024-05-22 13:56:28 -04:00
doctorpangloss
005e370254
Merge upstream
2024-03-21 13:15:36 -07:00
comfyanonymous
448d9263a2
Fix control loras breaking.
2024-03-14 09:30:21 -04:00
comfyanonymous
db8b59ecff
Lower memory usage for loras in lowvram mode at the cost of perf.
2024-03-13 20:07:27 -04:00
doctorpangloss
7520691021
Merge with master
2024-02-19 10:55:22 -08:00
comfyanonymous
667c92814e
Stable Cascade Stage B.
2024-02-16 13:02:03 -05:00
doctorpangloss
82edb2ff0e
Merge with latest upstream.
2024-01-29 15:06:31 -08:00
comfyanonymous
78a70fda87
Remove useless import.
2024-01-19 15:38:05 -05:00
comfyanonymous
36a7953142
Greatly improve lowvram sampling speed by getting rid of accelerate.
...
Let me know if this breaks anything.
2023-12-22 14:38:45 -05:00
comfyanonymous
77755ab8db
Refactor comfy.ops
...
comfy.ops -> comfy.ops.disable_weight_init
This should make it more clear what they actually do.
Some unused code has also been removed.
2023-12-11 23:27:13 -05:00
comfyanonymous
ba07cb748e
Use faster manual cast for fp8 in unet.
2023-12-11 18:24:44 -05:00
comfyanonymous
57926635e8
Switch text encoder to manual cast.
...
Use fp16 text encoder weights for CPU inference to lower memory usage.
2023-12-10 23:00:54 -05:00
comfyanonymous
af365e4dd1
All the unet ops with weights are now handled by comfy.ops
2023-12-04 03:12:18 -05:00
comfyanonymous
412d3ff57d
Refactor.
2023-11-11 01:11:06 -05:00
comfyanonymous
00c0b2c507
Initialize text encoder to target dtype.
2023-08-23 21:01:15 -04:00
comfyanonymous
d6e4b342e6
Support for Control Loras.
...
Control loras are controlnets where some of the weights are stored in
"lora" format: an up and a down low rank matrice that when multiplied
together and added to the unet weight give the controlnet weight.
This allows a much smaller memory footprint depending on the rank of the
matrices.
These controlnets are used just like regular ones.
2023-08-18 11:59:51 -04:00
comfyanonymous
bb1f45d6e8
Properly disable weight initialization in clip models.
2023-06-14 20:13:08 -04:00
comfyanonymous
21f04fe632
Disable default weight values in unet conv2d for faster loading.
2023-06-14 19:46:08 -04:00
comfyanonymous
6971646b8b
Speed up model loading a bit.
...
Default pytorch Linear initializes the weights which is useless and slow.
2023-06-14 12:09:41 -04:00
