Skip to content

mps: Alternative implementation for repeat_interleave #766

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.

Sign up for GitHub

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

Merged
pcuenca merged 9 commits into from
Oct 11, 2022
Merged

mps: Alternative implementation for repeat_interleave #766

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
46eb217
mps: alt. implementation for repeat_interleave
pcuenca Oct 7, 2022
54b02bb
style
pcuenca Oct 7, 2022
e4f753c
Bump mps version of PyTorch in the documentation.
pcuenca Oct 7, 2022
5fe1598
Apply suggestions from code review
patrickvonplaten Oct 7, 2022
f98fb14
Simplify: do not check for device.
pcuenca Oct 7, 2022
7bf3dec
style
pcuenca Oct 7, 2022
750d801
Merge remote-tracking branch 'origin/main' into mps-repeat-interleave
pcuenca Oct 11, 2022
c49bf48
Fix repeat dimensions:
pcuenca Oct 11, 2022
1aec2e5
Split long lines as suggested by Suraj.
pcuenca Oct 11, 2022
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
2 changes: 1 addition & 1 deletion docs/source/optimization/mps.mdx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@ specific language governing permissions and limitations under the License.
- Mac computer with Apple silicon (M1/M2) hardware.
- macOS 12.3 or later.
- arm64 version of Python.
- PyTorch [Preview (Nightly)](https://pytorch.org/get-started/locally/), version `1.13.0.dev20220830` or later.
- PyTorch [Preview (Nightly)](https://pytorch.org/get-started/locally/), version `1.14.0.dev20221007` or later.

## Inference Pipeline

Expand Down
12 changes: 8 additions & 4 deletions src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -218,8 +218,10 @@ def __call__(
text_input_ids = text_input_ids[:, : self.tokenizer.model_max_length]
text_embeddings = self.text_encoder(text_input_ids.to(self.device))[0]

# duplicate text embeddings for each generation per prompt
text_embeddings = text_embeddings.repeat_interleave(num_images_per_prompt, dim=0)
# duplicate text embeddings for each generation per prompt, using mps friendly method
bs_embed, seq_len, _ = text_embeddings.shape
text_embeddings = text_embeddings.repeat(1, num_images_per_prompt, 1)
text_embeddings = text_embeddings.view(bs_embed * num_images_per_prompt, seq_len, -1)

# here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
# of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`
Expand Down Expand Up @@ -256,8 +258,10 @@ def __call__(
)
uncond_embeddings = self.text_encoder(uncond_input.input_ids.to(self.device))[0]

# duplicate unconditional embeddings for each generation per prompt
uncond_embeddings = uncond_embeddings.repeat_interleave(batch_size * num_images_per_prompt, dim=0)
# duplicate unconditional embeddings for each generation per prompt, using mps friendly method
seq_len = uncond_embeddings.shape[1]
uncond_embeddings = uncond_embeddings.repeat(batch_size, num_images_per_prompt, 1)
uncond_embeddings = uncond_embeddings.view(batch_size * num_images_per_prompt, seq_len, -1)

# For classifier free guidance, we need to do two forward passes.
# Here we concatenate the unconditional and text embeddings into a single batch
Expand Down