Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

improvements #1

Open
wants to merge 98 commits into
base: master
Choose a base branch
from
Open
Changes from 1 commit
Commits
Show all changes
98 commits
Select commit Hold shift + click to select a range
0dccb46
Remove some unnecessary arguments.
comfyanonymous Jun 6, 2024
56333d4
Use the end token for the text encoder attention mask.
comfyanonymous Jun 7, 2024
6cd8ffc
Reshape the empty latent image to the right amount of channels if nee…
comfyanonymous Jun 8, 2024
742d572
Support zeroing out text embeddings with the attention mask.
comfyanonymous Jun 9, 2024
a5e6a63
Support sampling non 2D latents.
comfyanonymous Jun 10, 2024
8c4a9be
SD3 Support.
comfyanonymous Jun 10, 2024
a82fae2
Fix bug with cosxl edit model.
comfyanonymous Jun 10, 2024
4134564
Require safetensors library to be at least 0.4.2 for fp8 support.
comfyanonymous Jun 11, 2024
73ce178
Remove redundancy in mmdit.py (#3685)
Dango233 Jun 11, 2024
9424522
Reuse code.
comfyanonymous Jun 11, 2024
1c34d33
Update EmptySD3LatentImage to use 1024 resolution by default.
comfyanonymous Jun 11, 2024
5889b7c
Support multiple text encoder configurations on SD3.
comfyanonymous Jun 11, 2024
0e49211
Load the SD3 T5xxl model in the same dtype stored in the checkpoint.
comfyanonymous Jun 11, 2024
69c8d6d
Single and dual clip loader nodes support SD3.
comfyanonymous Jun 12, 2024
694e0b4
SD3 better memory usage estimation.
comfyanonymous Jun 12, 2024
32be358
Save SD3 modelspec.architecture in CheckpointSave node.
comfyanonymous Jun 12, 2024
1ddf512
Don't auto convert clip and vae weights to fp16 when saving checkpoint.
comfyanonymous Jun 12, 2024
c8b5e08
Default shift value on SD3 is 3.0
comfyanonymous Jun 12, 2024
321e509
Add link to SD3 example page to README.
comfyanonymous Jun 12, 2024
0eaa34e
Fix regular empty latent image not working with SD3 and custom sampler.
comfyanonymous Jun 12, 2024
605e64f
Fix lowvram issue.
comfyanonymous Jun 12, 2024
37a08a4
Support setting weight offsets in weight patcher.
comfyanonymous Jun 13, 2024
ac151ac
Support SD3 diffusers lora.
comfyanonymous Jun 13, 2024
5eb98f0
Exempt IPEX from non_blocking previews fixing segmentation faults. (#…
simonlui Jun 13, 2024
0e06b37
Print key names for easier debugging.
comfyanonymous Jun 14, 2024
0ec513d
Add a --force-channels-last to inference models in channel last mode.
comfyanonymous Jun 15, 2024
f2e844e
Optimize some unneeded if conditions in the sampling code.
comfyanonymous Jun 15, 2024
1281f93
Small optimization.
comfyanonymous Jun 15, 2024
bb1969c
Initial support for the stable audio open model.
comfyanonymous Jun 15, 2024
df7db0e
support TAESD3 (#3738)
ltdrdata Jun 16, 2024
04e8798
Improvements to the TAESD3 implementation.
comfyanonymous Jun 16, 2024
746a041
Fix VAEEncode with taesd3.
comfyanonymous Jun 16, 2024
ca9d300
Better estimation for memory usage during audio VAE encoding/decoding.
comfyanonymous Jun 16, 2024
8ddc151
Squash depreciation warning on new pytorch.
comfyanonymous Jun 16, 2024
6425252
Use fp16 as the default vae dtype for the audio VAE.
comfyanonymous Jun 16, 2024
b7c473d
Fix lora keys for SimpleTuner (#3759)
janekm Jun 17, 2024
379ff92
fix app.js no graph defined (#3754)
o0OooO Jun 17, 2024
a45df69
Basic tiled decoding for audio VAE.
comfyanonymous Jun 18, 2024
55f0dc1
Add soundfile dependency so that windows can save audio.
comfyanonymous Jun 18, 2024
3914d5a
Support full SD3 loras.
comfyanonymous Jun 19, 2024
97ae6ef
Add api/ prefix to api endpoints (#3779)
huchenlei Jun 19, 2024
e11052a
Add ipndm sampler.
comfyanonymous Jun 19, 2024
eee815e
Update sd1_clip.py (#3684)
Quasimondo Jun 19, 2024
b08a9dd
Remove empty line.
comfyanonymous Jun 20, 2024
0d6a579
Support loading diffusers SD3 model format with UNETLoader node.
comfyanonymous Jun 20, 2024
028a583
Fix issue with full diffusers SD3 loras.
comfyanonymous Jun 20, 2024
d7f0964
Fix routes (#3790)
huchenlei Jun 20, 2024
45e10ca
feat: add gits scheduler (#3769)
zhyzhouu Jun 20, 2024
d5efde8
Add ipndm_v sampler, works best with the exponential scheduler.
comfyanonymous Jun 20, 2024
1e2839f
More proper tiled audio decoding.
comfyanonymous Jun 20, 2024
887a634
Proper ModelMergeSD3_2B node.
comfyanonymous Jun 21, 2024
4ef1479
Multi dimension tiled scale function and tiled VAE audio encoding fal…
comfyanonymous Jun 22, 2024
2f360ae
Support OneTrainer SD3 lora format.
comfyanonymous Jun 22, 2024
73ca780
Add SamplerEulerCFG++ node.
comfyanonymous Jun 23, 2024
866f54d
Add browser test action synced with TS repo (#3852)
huchenlei Jun 24, 2024
eab211b
Resample audio to 44100 when VAE encoding it.
comfyanonymous Jun 24, 2024
90aebb6
New Menu & Workflow Management (#3112)
pythongosssss Jun 25, 2024
69d710e
Implement my alternative take on CFG++ as the euler_pp sampler.
comfyanonymous Jun 25, 2024
e99d97a
Remove duplicated Reset View button (#3865)
huchenlei Jun 26, 2024
edfce78
add issue templates for ComfyUI Issues Page (#3868)
mcmonkey4eva Jun 26, 2024
e3579f3
Fix merge issue breaking api json loading (#3876)
pythongosssss Jun 26, 2024
f12fa1d
Enable browser tests on push (#3878)
huchenlei Jun 26, 2024
a3e83f6
Update test ref (#3882)
huchenlei Jun 26, 2024
bc5a0f1
Ignore *.log (#3880)
huchenlei Jun 26, 2024
175fe02
Ignore .vscode/ (#3879)
huchenlei Jun 26, 2024
44947e7
Add DEIS order 3 sampler.
comfyanonymous Jun 27, 2024
3b423af
Add audio widget (#3863)
huchenlei Jun 27, 2024
4650e7d
Save and load workflow from the flac files output by SaveAudio.
comfyanonymous Jun 27, 2024
4f9d2b0
Remove print.
comfyanonymous Jun 27, 2024
5ff3d4e
Fix audio upload when no audio in input dir (#3891)
huchenlei Jun 27, 2024
8ceb5a0
Support saving stable audio checkpoint that can be loaded back.
comfyanonymous Jun 27, 2024
449bf52
Add integration test for Linux with Nvidia GPU. #3884 (#3895)
robinjhuang Jun 27, 2024
97b409c
Add macOs integration test for default workflow. (#3898)
robinjhuang Jun 27, 2024
66aaa14
Controlnet refactor.
comfyanonymous Jun 25, 2024
f8f7568
Basic SD3 controlnet implementation.
comfyanonymous Jun 26, 2024
264caca
ControlNetApplySD3 node can now be used to use SD3 controlnets.
comfyanonymous Jun 26, 2024
0d9009c
New menu/workflows fixes (#3900)
pythongosssss Jun 28, 2024
7ecb2ec
Audio second setting in EmptyLatentAudio.
comfyanonymous Jun 28, 2024
02cac1d
Revert "Add macOs integration test for default workflow. (#3898)" (#3…
robinjhuang Jun 28, 2024
c39cf7f
Revert "Add integration test for Linux with Nvidia GPU. #3884 (#3895)…
robinjhuang Jun 28, 2024
fbb7a1f
PreviewAudio node.
comfyanonymous Jun 29, 2024
05e8316
Switch to the real cfg++ method in the samplers.
comfyanonymous Jun 29, 2024
dbb7dd3
Add to readme that Stable Audio is supported.
comfyanonymous Jun 30, 2024
521421f
Fix workflow not importing from flac files on some systems.
comfyanonymous Jun 30, 2024
7c5fa7f
Fix loadGraphData func call (#3918)
huchenlei Jul 1, 2024
e53b159
enable cmd shortcuts for mac (mute & bypass) (#3792)
ruucm Jul 1, 2024
601b4b6
Add CONTRIBUTING.md (#3910)
robinjhuang Jul 1, 2024
0cd4a6a
Fine-tuning GitHub Actions (#3169)
YanWenKun Jul 1, 2024
5dccfef
Switch nightly pytorch standalone package to lzma2.
comfyanonymous Jul 1, 2024
755c48d
Fix several typos in example_node.py.example (#3204)
HaydenReeve Jul 1, 2024
b82d67d
Add SamplerEulerAncestralCFG++ custom sampler node (#3901)
petercrabtree Jul 1, 2024
9dd549e
Add `--no-custom-node` cmd flag (#3903)
huchenlei Jul 1, 2024
1ef66b0
Add example for how to add custom API routes (#3597)
BobDu Jul 1, 2024
52aaee2
Fix to #3465. Prevent, resaving of duplicate images if overwrite not …
shawnington Jul 2, 2024
2f03201
Remove some empty lines.
comfyanonymous Jul 2, 2024
60bb105
Print node execution times to the log
rdancer Jun 15, 2024
7c6b3eb
fix ImageBlend for batches
rdancer Jun 18, 2024
64588bb
add debugging stub to ComfyUI startup
rdancer Jul 7, 2024
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
Small optimization.
  • Loading branch information
comfyanonymous committed Jun 15, 2024
commit 1281f933c1c38ac0491ff2f86cbcd2ec90743ce3
4 changes: 2 additions & 2 deletions comfy/ldm/modules/diffusionmodules/mmdit.py
Original file line number Diff line number Diff line change
Expand Up @@ -243,9 +243,9 @@ def timestep_embedding(t, dim, max_period=10000):
half = dim // 2
freqs = torch.exp(
-math.log(max_period)
* torch.arange(start=0, end=half, dtype=torch.float32)
* torch.arange(start=0, end=half, dtype=torch.float32, device=t.device)
/ half
).to(device=t.device)
)
args = t[:, None].float() * freqs[None]
embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
if dim % 2:
Expand Down