KenSchae
/
Pallaidium
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
Browse Source

Attempt to fix CUDA and other errors.

Inpaint_experimental
tin2tin 1 year ago committed by GitHub
parent
2a04a52668
commit
a5ffd03bc1
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 42 additions and 38 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 80
      __init__.py

80
__init__.py
Unescape View File

@ -391,7 +391,7 @@ def low_vram():
for i in range(torch.cuda.device_count()):
properties = torch.cuda.get_device_properties(i)
total_vram += properties.total_memory
return (total_vram / (1024**3)) < 6.1 # Y/N under 6.1 GB?
return (total_vram / (1024**3)) < 8.1 # Y/N under 6.1 GB?
def import_module(self, module, install_module):
@ -500,10 +500,10 @@ def install_modules(self):
import_module(self, "sox", "sox")
else:
import_module(self, "soundfile", "PySoundFile")
import_module(self, "diffusers", "diffusers")
#import_module(self, "diffusers", "diffusers")
import_module(self, "diffusers", "git+https://github.com/huggingface/diffusers.git@v0.19.3")
# import_module(self, "diffusers", "git+https://github.com/huggingface/diffusers.git")
import_module(self, "accelerate", "accelerate")
# import_module(self, "diffusers", "git+https://github.com/huggingface/accelerate.git")
import_module(self, "transformers", "transformers")
# import_module(self, "optimum", "optimum")
import_module(self, "sentencepiece", "sentencepiece")
@ -1006,6 +1006,7 @@ class SEQUENCER_OT_generate_movie(Operator):
import torch
from diffusers import (
DiffusionPipeline,
StableDiffusionXLPipeline,
DPMSolverMultistepScheduler,
TextToVideoSDPipeline,
VideoToVideoSDPipeline,
@ -1065,8 +1066,9 @@ class SEQUENCER_OT_generate_movie(Operator):
import torch
from diffusers import StableDiffusionXLImg2ImgPipeline
pipe = DiffusionPipeline.from_pretrained(
image_model_card,
pipe = StableDiffusionXLPipeline.from_pretrained(
"stabilityai/stable-diffusion-xl-base-1.0",
#image_model_card,
torch_dtype=torch.float16,
variant="fp16",
)
@ -1079,9 +1081,9 @@ class SEQUENCER_OT_generate_movie(Operator):
torch.cuda.set_per_process_memory_fraction(0.95)
pipe.enable_model_cpu_offload()
# pipe.unet.enable_forward_chunking(chunk_size=1, dim=1)
pipe.unet.added_cond_kwargs={}
#pipe.unet.added_cond_kwargs={}
pipe.enable_vae_slicing()
pipe.enable_xformers_memory_efficient_attention()
#pipe.enable_xformers_memory_efficient_attention()
else:
pipe.to("cuda")
@ -1101,19 +1103,21 @@ class SEQUENCER_OT_generate_movie(Operator):
torch.cuda.set_per_process_memory_fraction(0.95)
refiner.enable_model_cpu_offload()
# refiner.unet.enable_forward_chunking(chunk_size=1, dim=1)
refiner.unet.added_cond_kwargs={}
#refiner.unet.added_cond_kwargs={}
refiner.enable_vae_slicing()
refiner.enable_xformers_memory_efficient_attention()
#refiner.enable_xformers_memory_efficient_attention()
else:
refiner.to("cuda")
else:
if movie_model_card == "cerspense/zeroscope_v2_dark_30x448x256" or movie_model_card == "cerspense/zeroscope_v2_576w":
card = "stabilityai/stable-diffusion-xl-base-1.0"
else:
card = movie_model_card
# if movie_model_card == "cerspense/zeroscope_v2_dark_30x448x256" or movie_model_card == "cerspense/zeroscope_v2_576w":
# card = "stabilityai/stable-diffusion-xl-base-1.0"
# else:
# card = movie_model_card
upscale = VideoToVideoSDPipeline.from_pretrained(
card,
# "cerspense/zeroscope_v2_576w",
"cerspense/zeroscope_v2_XL",
torch_dtype=torch.float16,
#text_encoder=upscale.text_encoder,
#vae=upscale.vae,
@ -1123,20 +1127,21 @@ class SEQUENCER_OT_generate_movie(Operator):
upscale.scheduler = DPMSolverMultistepScheduler.from_config(upscale.scheduler.config)
if low_vram:
torch.cuda.set_per_process_memory_fraction(0.95) # 6 GB VRAM
# torch.cuda.set_per_process_memory_fraction(0.95) # 6 GB VRAM
upscale.enable_model_cpu_offload()
# upscale.unet.enable_forward_chunking(chunk_size=1, dim=1)
upscale.unet.added_cond_kwargs={}
upscale.unet.enable_forward_chunking(chunk_size=1, dim=1)
#upscale.unet.added_cond_kwargs={}
upscale.enable_vae_slicing()
upscale.enable_xformers_memory_efficient_attention()
#upscale.enable_xformers_memory_efficient_attention()
else:
upscale.to("cuda")
# Models for movie generation
else:
# Options: https://huggingface.co/docs/diffusers/api/pipelines/text_to_video
pipe = TextToVideoSDPipeline.from_pretrained(
#pipe = TextToVideoSDPipeline.from_pretrained(
pipe = DiffusionPipeline.from_pretrained(
movie_model_card,
torch_dtype=torch.float16,
# variant="fp16",
@ -1149,33 +1154,35 @@ class SEQUENCER_OT_generate_movie(Operator):
if low_vram:
pipe.enable_model_cpu_offload()
# pipe.unet.enable_forward_chunking(chunk_size=1, dim=1)
pipe.unet.added_cond_kwargs={}
#pipe.unet.added_cond_kwargs={}
pipe.enable_vae_slicing()
pipe.enable_xformers_memory_efficient_attention()
#pipe.enable_xformers_memory_efficient_attention()
else:
pipe.to("cuda")
# Models for upscale generated movie
# Model for upscale generated movie
if scene.video_to_video:
if torch.cuda.is_available():
torch.cuda.empty_cache()
# torch.cuda.set_per_process_memory_fraction(0.85) # 6 GB VRAM
upscale = VideoToVideoSDPipeline.from_pretrained(
# upscale = VideoToVideoSDPipeline.from_pretrained(
upscale = DiffusionPipeline.from_pretrained(
#"cerspense/zeroscope_v2_576w", torch_dtype=torch.float16
"cerspense/zeroscope_v2_XL", torch_dtype=torch.float16
)
# upscale = VideoToVideoSDPipeline.from_pretrained("cerspense/zeroscope_v2_576w", torch_dtype=torch.float16)
upscale.scheduler = DPMSolverMultistepScheduler.from_config(
pipe.scheduler.config
upscale.scheduler.config
)
if low_vram:
upscale.enable_model_cpu_offload()
# upscale.unet.enable_forward_chunking(chunk_size=1, dim=1)
upscale.unet.added_cond_kwargs={}
upscale.unet.enable_forward_chunking(chunk_size=1, dim=1)
#upscale.unet.added_cond_kwargs={}
upscale.enable_vae_slicing()
upscale.enable_xformers_memory_efficient_attention()
#upscale.enable_xformers_memory_efficient_attention()
else:
upscale.to("cuda")
@ -1275,7 +1282,7 @@ class SEQUENCER_OT_generate_movie(Operator):
# Upscale video
if scene.video_to_video:
video = [
Image.fromarray(frame).resize((int(x * 2), int(y * 2)))
Image.fromarray(frame).resize((closest_divisible_64(int(x * 2)), closest_divisible_64(int(y * 2))))
for frame in video
]
@ -1313,10 +1320,7 @@ class SEQUENCER_OT_generate_movie(Operator):
print("Upscale Video")
if torch.cuda.is_available():
torch.cuda.empty_cache()
video = [
Image.fromarray(frame).resize((x * 2, y * 2))
for frame in video_frames
]
video = [Image.fromarray(frame).resize((closest_divisible_64(x * 2), closest_divisible_64(y * 2))) for frame in video_frames]
video_frames = upscale(
prompt,
@ -1465,9 +1469,9 @@ class SEQUENCER_OT_generate_audio(Operator):
if low_vram:
pipe.enable_model_cpu_offload()
# pipe.unet.enable_forward_chunking(chunk_size=1, dim=1)
pipe.unet.added_cond_kwargs={}
# pipe.unet.added_cond_kwargs={}
pipe.enable_vae_slicing()
pipe.enable_xformers_memory_efficient_attention()
#pipe.enable_xformers_memory_efficient_attention()
else:
pipe.to("cuda")
@ -1671,9 +1675,9 @@ class SEQUENCER_OT_generate_image(Operator):
torch.cuda.set_per_process_memory_fraction(0.95) # 6 GB VRAM
pipe.enable_model_cpu_offload()
# pipe.unet.enable_forward_chunking(chunk_size=1, dim=1)
pipe.unet.added_cond_kwargs={}
#pipe.unet.added_cond_kwargs={}
pipe.enable_vae_slicing()
pipe.enable_xformers_memory_efficient_attention()
#pipe.enable_xformers_memory_efficient_attention()
else:
pipe.to("cuda")
@ -1745,9 +1749,9 @@ class SEQUENCER_OT_generate_image(Operator):
if low_vram:
refiner.enable_model_cpu_offload()
# refiner.unet.enable_forward_chunking(chunk_size=1, dim=1)
refiner.unet.added_cond_kwargs={}
#refiner.unet.added_cond_kwargs={}
refiner.enable_vae_slicing()
refiner.enable_xformers_memory_efficient_attention()
#refiner.enable_xformers_memory_efficient_attention()
else:
refiner.to("cuda")

Write Preview
Loading…
Cancel
Save