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

Refinement for gpu settings #119

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Refinement for gpu settings #119

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
90601b8
Update webgui.py for gpu settings
CodeLyokoscj Aug 5, 2024
a4e9973
Update mutual_self_attention.py for gpu settings
CodeLyokoscj Aug 5, 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
3 changes: 0 additions & 3 deletions src/models/mutual_self_attention.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,13 +80,11 @@ def register_reference_hooks(
[1] * batch_size * num_images_per_prompt * 16
+ [0] * batch_size * num_images_per_prompt * 16
)
.to(device)
.bool()
)
else:
uc_mask = (
torch.Tensor([0] * batch_size * num_images_per_prompt * 2)
.to(device)
.bool()
)

Expand Down Expand Up @@ -170,7 +168,6 @@ def hacked_basic_transformer_inner_forward(
[1] * (hidden_states.shape[0] // 2)
+ [0] * (hidden_states.shape[0] // 2)
)
.to(device)
.bool()
)
hidden_states_c[_uc_mask] = (
Expand Down
8 changes: 4 additions & 4 deletions webgui.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,7 +67,7 @@

############# model_init started #############
## vae init
vae = AutoencoderKL.from_pretrained(config.pretrained_vae_path).to("cuda", dtype=weight_dtype)
vae = AutoencoderKL.from_pretrained(config.pretrained_vae_path).to(device, dtype=weight_dtype)

## reference net init
reference_unet = UNet2DConditionModel.from_pretrained(
Expand Down Expand Up @@ -101,7 +101,7 @@
denoising_unet.load_state_dict(torch.load(config.denoising_unet_path, map_location="cpu"), strict=False)

## face locator init
face_locator = FaceLocator(320, conditioning_channels=1, block_out_channels=(16, 32, 96, 256)).to(dtype=weight_dtype, device="cuda")
face_locator = FaceLocator(320, conditioning_channels=1, block_out_channels=(16, 32, 96, 256)).to(dtype=weight_dtype, device=device)
face_locator.load_state_dict(torch.load(config.face_locator_path))

## load audio processor params
Expand All @@ -122,7 +122,7 @@
audio_guider=audio_processor,
face_locator=face_locator,
scheduler=scheduler,
).to("cuda", dtype=weight_dtype)
).to(device, dtype=weight_dtype)

def select_face(det_bboxes, probs):
## max face from faces that the prob is above 0.8
Expand Down Expand Up @@ -170,7 +170,7 @@ def process_video(uploaded_img, uploaded_audio, width, height, length, seed, fac
face_mask = cv2.resize(face_mask, (width, height))

ref_image_pil = Image.fromarray(face_img[:, :, [2, 1, 0]])
face_mask_tensor = torch.Tensor(face_mask).to(dtype=weight_dtype, device="cuda").unsqueeze(0).unsqueeze(0).unsqueeze(0) / 255.0
face_mask_tensor = torch.Tensor(face_mask).to(dtype=weight_dtype, device=device).unsqueeze(0).unsqueeze(0).unsqueeze(0) / 255.0

video = pipe(
ref_image_pil,
Expand Down