Skip to content

Fixed fused_memory_opt type interference error #8

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

Open
Maximilian-Winter wants to merge 8 commits into
base: main
Choose a base branch
from
Open

Fixed fused_memory_opt type interference error #8

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
318650e
Fixed fused_memory_opt error
Maximilian-Winter Sep 19, 2022
5f7c6c4
Fixed requirements and save as single image with time stamp
Maximilian-Winter Sep 20, 2022
3813301
Merge https://github.com/neonsecret/stable-diffusion
Maximilian-Winter Sep 20, 2022
dcb6ec7
Added pngs to git ignore
Maximilian-Winter Sep 20, 2022
1a4f43a
Test
Maximilian-Winter Sep 20, 2022
da7c316
Test2
Maximilian-Winter Sep 20, 2022
825de2e
Merge https://github.com/neonsecret/stable-diffusion
Maximilian-Winter Sep 20, 2022
f47062c
Merge https://github.com/neonsecret/stable-diffusion
Maximilian-Winter Sep 30, 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
16 changes: 16 additions & 0 deletions .gitignore
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,16 @@

*.png
*.pyc
*.csv
CodeFormer/
src/clip/
src/taming-transformers/
latent_diffusion.egg-info/dependency_links.txt
latent_diffusion.egg-info/PKG-INFO
latent_diffusion.egg-info/requires.txt
latent_diffusion.egg-info/SOURCES.txt
latent_diffusion.egg-info/top_level.txt
log.txt
models/ldm/stable-diffusion-v1/model.ckpt
.vscode/settings.json
tqdm.txt
4 changes: 2 additions & 2 deletions ldm/modules/diffusionmodules/model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -184,7 +184,7 @@ def __init__(self, in_channels):


@torch.jit.script
def fused_t(x, c):
def fused_t(x, c: int):
return x * (int(c) ** (-0.5))


Expand All @@ -197,7 +197,7 @@ def fused_memory_opt(mem_reserved, mem_active, mem_free_cuda, b, h, w, c):
# s3 = (b * c * h * w * 3) * 2 # zeros_like, empty_like, empty_strided
# s = 2 * (s1 + s2 + s3) # 2 because of small allocations which don't really matter
s = 16 * b * ((h * w) ** 2) + 12 * b * c * h * w
s = (s // mem_free_total) + 1 if s > mem_free_total else torch.tensor(1)
s = int((s // mem_free_total) + 1) if s > mem_free_total else 1
return s


Expand Down
20 changes: 17 additions & 3 deletions optimizedSD/optimized_txt2img.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,5 @@
import argparse
from datetime import datetime
import os
import random
import re
Expand Down Expand Up @@ -117,6 +118,7 @@ def get_image(opt, model, modelCS, modelFS, prompt=None, save=True, callback_fn=
negative_prompt = ""
with torch.no_grad():
all_samples = list()
count_iter = 0
for _ in trange(opt.n_iter, desc="Sampling"):
for prompts in tqdm(data, desc="data"):
sample_path = os.path.join(opt.outpath, "_".join(re.split(":| ", prompts[0])))[:150]
Expand Down Expand Up @@ -147,7 +149,7 @@ def get_image(opt, model, modelCS, modelFS, prompt=None, save=True, callback_fn=
c = modelCS.get_learned_conditioning(prompts)

shape = [opt.num_images, opt.C, opt.height // opt.f, opt.width // opt.f]

seed_at_begin = opt.seed
if opt.device != "cpu":
mem = torch.cuda.memory_allocated() / 1e6
modelCS.to("cpu")
Expand Down Expand Up @@ -190,14 +192,23 @@ def get_image(opt, model, modelCS, modelFS, prompt=None, save=True, callback_fn=
opt.seed += 1
base_count += 1

count = 0
for x_index in range(count_iter * batch_size, len(all_samples)):
x_sample = 255. * rearrange(all_samples[x_index][0].numpy(), 'c h w -> h w c')
now = datetime.now()
dt_string = now.strftime("%d-%m-%Y %H_%M_%S ")
Image.fromarray(x_sample.astype(np.uint8)).save(
os.path.join(sample_path, f"{dt_string}s{seed_at_begin}d{opt.ddim_steps}-{str(count)}.png"))
count += 1

if opt.device != "cpu":
mem = torch.cuda.memory_allocated() / 1e6
modelFS.to("cpu")
while torch.cuda.memory_allocated() / 1e6 >= mem:
time.sleep(1)
del samples_ddim
print("memory_final = ", torch.cuda.memory_allocated() / 1e6)

count_iter += 1
toc = time.time()

time_taken = (toc - tic) / 60.0
Expand Down Expand Up @@ -430,7 +441,10 @@ def get_image(opt, model, modelCS, modelFS, prompt=None, save=True, callback_fn=
grid = torch.cat(all_samples, 0)
grid = make_grid(grid, nrow=opt.n_iter)
grid = 255.0 * rearrange(grid, "c h w -> h w c").cpu().numpy()

now = datetime.now()
dt_string = now.strftime(" %d-%m-%Y %H_%M_%S ")
Image.fromarray(grid.astype(np.uint8)).save(
os.path.join(outpath + "/" + str(opt.prompt).replace("/", "")[:100] + f".{opt.format}")
os.path.join(outpath + "/" + str(opt.prompt).replace("/", "")[:100] + dt_string + f".{opt.format}")
)
print("exported to", outpath + "/" + str(opt.prompt).replace("/", "")[:100] + f".{opt.format}")