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.

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

Jump to bottom

Support Metal Performance Shaders #125

Open
wants to merge 7 commits into
base: main
Choose a base branch
from
Open

Support Metal Performance Shaders #125

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
f164e36
add mps choice
Tps-F Feb 17, 2024
6699b70
support mps event
Tps-F Feb 17, 2024
1b66259
timerevent in benchmark
Tps-F Feb 17, 2024
16baeb5
using getattr to write simple
Tps-F Feb 17, 2024
22c5c59
fix device
Tps-F Feb 17, 2024
70c523e
skip acceleration on mps
Tps-F Feb 17, 2024
d579e64
Merge branch 'cumulo-autumn:main' into main
Tps-F Mar 4, 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
7 changes: 4 additions & 3 deletions examples/benchmark/multi.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -127,15 +127,16 @@ def run(

results = []

start = torch.cuda.Event(enable_timing=True)
end = torch.cuda.Event(enable_timing=True)
timer_event = getattr(torch, "cuda" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu")
start = timer_event.Event(enable_timing=True)
end = timer_event.Event(enable_timing=True)
for _ in tqdm(range(iterations)):
start.record()
out_tensor = stream.stream(image_tensor).cpu()
queue.put(out_tensor)
end.record()

torch.cuda.synchronize()
timer_event.synchronize()
results.append(start.elapsed_time(end))

print(f"Average time: {sum(results) / len(results)}ms")
Expand Down
7 changes: 4 additions & 3 deletions examples/benchmark/single.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -112,16 +112,17 @@ def run(

results = []

start = torch.cuda.Event(enable_timing=True)
end = torch.cuda.Event(enable_timing=True)
timer_event = getattr(torch, "cuda" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu")
start = timer_event.Event(enable_timing=True)
end = timer_event.Event(enable_timing=True)

for _ in tqdm(range(iterations)):
start.record()
image_tensor = stream.preprocess_image(downloaded_image)
stream(image=image_tensor)
end.record()

torch.cuda.synchronize()
timer_event.synchronize()
results.append(start.elapsed_time(end))

print(f"Average time: {sum(results) / len(results)}ms")
Expand Down
9 changes: 6 additions & 3 deletions src/streamdiffusion/pipeline.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,5 @@
import time

from typing import List, Optional, Union, Any, Dict, Tuple, Literal

import numpy as np
Expand Down Expand Up @@ -30,6 +31,8 @@ def __init__(
self.dtype = torch_dtype
self.generator = None

self.timer_event = getattr(torch, str(self.device).split(':', 1)[0])

self.height = height
self.width = width

Expand Down Expand Up @@ -440,8 +443,8 @@ def predict_x0_batch(self, x_t_latent: torch.Tensor) -> torch.Tensor:
def __call__(
self, x: Union[torch.Tensor, PIL.Image.Image, np.ndarray] = None
) -> torch.Tensor:
start = torch.cuda.Event(enable_timing=True)
end = torch.cuda.Event(enable_timing=True)
start = self.timer_event.Event(enable_timing=True)
end = self.timer_event.Event(enable_timing=True)
start.record()
if x is not None:
x = self.image_processor.preprocess(x, self.height, self.width).to(
Expand All @@ -463,7 +466,7 @@ def __call__(

self.prev_image_result = x_output
end.record()
torch.cuda.synchronize()
self.timer_event.synchronize()
inference_time = start.elapsed_time(end) / 1000
self.inference_time_ema = 0.9 * self.inference_time_ema + 0.1 * inference_time
return x_output
Expand Down
Loading