[feat] allow SDXL pipeline to run with fused QKV projections #6030
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
https://github.com/sayakpaul/sdxl-fast Also, note that this is just one of the many things needed to speed things up. But throughput-wise it does contribute and one can check with the following code: from diffusers import DiffusionPipeline
import torch
pipeline = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16).to("cuda")
pipeline.fuse_qkv_projections()
for _ in range(5):
_ = pipeline("hey", num_inference_steps=25).images[0]
pipeline.unfuse_qkv_projections()
for _ in range(5):
_ = pipeline("hey", num_inference_steps=25).images[0]You should see speedup in the throughput. |
Co-authored-by: Patrick von Platen <[email protected]>
|
Indeed, getting a nice 5% speed-up! |
src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py
Outdated
Show resolved
Hide resolved
src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py
Outdated
Show resolved
Hide resolved
Co-authored-by: Patrick von Platen <[email protected]>
|
@DN6 any reason why the fetcher is failing? |
patrickvonplaten
approved these changes
Dec 5, 2023
| @@ -289,6 +290,8 @@ def set_timesteps(self, num_inference_steps: int, device: Union[str, torch.devic | |||
| self.timesteps = torch.from_numpy(timesteps.astype(np.float32)).to(device=device) | |||
|
|
|||
| self.sigmas = torch.cat([sigmas, torch.zeros(1, device=sigmas.device)]) | |||
| if sigmas.device.type == "cuda": | |||
There was a problem hiding this comment.
this breaks the add_noise function and thus inpaint and img2img and training
donhardman
pushed a commit
to donhardman/diffusers
that referenced
this pull request
Dec 18, 2023
…face#6030) * debug * from step * print * turn sigma a list * make str * init_noise_sigma * comment * remove prints * feat: introduce fused projections * change to a better name * no grad * device. * device * dtype * okay * print * more print * fix: unbind -> split * fix: qkv >-> k * enable disable * apply attention processor within the method * attn processors * _enable_fused_qkv_projections * remove print * add fused projection to vae * add todos. * add: documentation and cleanups. * add: test for qkv projection fusion. * relax assertions. * relax further * fix: docs * fix-copies * correct error message. * Empty-Commit * better conditioning on disable_fused_qkv_projections * check * check processor * bfloat16 computation. * check latent dtype * style * remove copy temporarily * cast latent to bfloat16 * fix: vae -> self.vae * remove print. * add _change_to_group_norm_32 * comment out stuff that didn't work * Apply suggestions from code review Co-authored-by: Patrick von Platen <[email protected]> * reflect patrick's suggestions. * fix imports * fix: disable call. * fix more * fix device and dtype * fix conditions. * fix more * Apply suggestions from code review Co-authored-by: Patrick von Platen <[email protected]> --------- Co-authored-by: Patrick von Platen <[email protected]>
yoonseokjin
pushed a commit
to yoonseokjin/diffusers
that referenced
this pull request
Dec 25, 2023
…face#6030) * debug * from step * print * turn sigma a list * make str * init_noise_sigma * comment * remove prints * feat: introduce fused projections * change to a better name * no grad * device. * device * dtype * okay * print * more print * fix: unbind -> split * fix: qkv >-> k * enable disable * apply attention processor within the method * attn processors * _enable_fused_qkv_projections * remove print * add fused projection to vae * add todos. * add: documentation and cleanups. * add: test for qkv projection fusion. * relax assertions. * relax further * fix: docs * fix-copies * correct error message. * Empty-Commit * better conditioning on disable_fused_qkv_projections * check * check processor * bfloat16 computation. * check latent dtype * style * remove copy temporarily * cast latent to bfloat16 * fix: vae -> self.vae * remove print. * add _change_to_group_norm_32 * comment out stuff that didn't work * Apply suggestions from code review Co-authored-by: Patrick von Platen <[email protected]> * reflect patrick's suggestions. * fix imports * fix: disable call. * fix more * fix device and dtype * fix conditions. * fix more * Apply suggestions from code review Co-authored-by: Patrick von Platen <[email protected]> --------- Co-authored-by: Patrick von Platen <[email protected]>
AmericanPresidentJimmyCarter
pushed a commit
to AmericanPresidentJimmyCarter/diffusers
that referenced
this pull request
Apr 26, 2024
…face#6030) * debug * from step * print * turn sigma a list * make str * init_noise_sigma * comment * remove prints * feat: introduce fused projections * change to a better name * no grad * device. * device * dtype * okay * print * more print * fix: unbind -> split * fix: qkv >-> k * enable disable * apply attention processor within the method * attn processors * _enable_fused_qkv_projections * remove print * add fused projection to vae * add todos. * add: documentation and cleanups. * add: test for qkv projection fusion. * relax assertions. * relax further * fix: docs * fix-copies * correct error message. * Empty-Commit * better conditioning on disable_fused_qkv_projections * check * check processor * bfloat16 computation. * check latent dtype * style * remove copy temporarily * cast latent to bfloat16 * fix: vae -> self.vae * remove print. * add _change_to_group_norm_32 * comment out stuff that didn't work * Apply suggestions from code review Co-authored-by: Patrick von Platen <[email protected]> * reflect patrick's suggestions. * fix imports * fix: disable call. * fix more * fix device and dtype * fix conditions. * fix more * Apply suggestions from code review Co-authored-by: Patrick von Platen <[email protected]> --------- Co-authored-by: Patrick von Platen <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do?
Adds an option for running the SDXL pipeline with QKV projections fused. For self-attention, all the projection matrices are horizontally fused. For cross-attention, key and value projection matrices are fused.
Some more comments are inline.
A lot of thanks to @cpuhrsch for helping.