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

Fix MPS usage via autocast change #84

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

Fix MPS usage via autocast change #84

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
642a981
Update autocast to use CPU for MPS type
bcherry Jun 26, 2023
2139232
Delete test.py
bcherry Jun 27, 2023
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
12 changes: 6 additions & 6 deletions clip_interrogator/clip_interrogator.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -197,7 +197,7 @@ def generate_caption(self, pil_image: Image) -> str:
def image_to_features(self, image: Image) -> torch.Tensor:
self._prepare_clip()
images = self.clip_preprocess(image).unsqueeze(0).to(self.device)
with torch.no_grad(), torch.cuda.amp.autocast():
with torch.no_grad(), torch.amp.autocast(device_type='cuda' if self.device == 'cuda' else 'cpu'):
image_features = self.clip_model.encode_image(images)
image_features /= image_features.norm(dim=-1, keepdim=True)
return image_features
Expand Down Expand Up @@ -257,7 +257,7 @@ def interrogate(self, image: Image, min_flavors: int=8, max_flavors: int=32, cap
def rank_top(self, image_features: torch.Tensor, text_array: List[str], reverse: bool=False) -> str:
self._prepare_clip()
text_tokens = self.tokenize([text for text in text_array]).to(self.device)
with torch.no_grad(), torch.cuda.amp.autocast():
with torch.no_grad(), torch.amp.autocast(device_type='cuda' if self.device == 'cuda' else 'cpu'):
text_features = self.clip_model.encode_text(text_tokens)
text_features /= text_features.norm(dim=-1, keepdim=True)
similarity = text_features @ image_features.T
Expand All @@ -268,7 +268,7 @@ def rank_top(self, image_features: torch.Tensor, text_array: List[str], reverse:
def similarity(self, image_features: torch.Tensor, text: str) -> float:
self._prepare_clip()
text_tokens = self.tokenize([text]).to(self.device)
with torch.no_grad(), torch.cuda.amp.autocast():
with torch.no_grad(), torch.amp.autocast(device_type='cuda' if self.device == 'cuda' else 'cpu'):
text_features = self.clip_model.encode_text(text_tokens)
text_features /= text_features.norm(dim=-1, keepdim=True)
similarity = text_features @ image_features.T
Expand All @@ -277,7 +277,7 @@ def similarity(self, image_features: torch.Tensor, text: str) -> float:
def similarities(self, image_features: torch.Tensor, text_array: List[str]) -> List[float]:
self._prepare_clip()
text_tokens = self.tokenize([text for text in text_array]).to(self.device)
with torch.no_grad(), torch.cuda.amp.autocast():
with torch.no_grad(), torch.amp.autocast(device_type='cuda' if self.device == 'cuda' else 'cpu'):
text_features = self.clip_model.encode_text(text_tokens)
text_features /= text_features.norm(dim=-1, keepdim=True)
similarity = text_features @ image_features.T
Expand Down Expand Up @@ -319,7 +319,7 @@ def __init__(self, labels:List[str], desc:str, ci: Interrogator):
chunks = np.array_split(self.labels, max(1, len(self.labels)/config.chunk_size))
for chunk in tqdm(chunks, desc=f"Preprocessing {desc}" if desc else None, disable=self.config.quiet):
text_tokens = self.tokenize(chunk).to(self.device)
with torch.no_grad(), torch.cuda.amp.autocast():
with torch.no_grad(), torch.amp.autocast(device_type='cuda' if self.device == 'cuda' else 'cpu'):
text_features = clip_model.encode_text(text_tokens)
text_features /= text_features.norm(dim=-1, keepdim=True)
text_features = text_features.half().cpu().numpy()
Expand Down Expand Up @@ -373,7 +373,7 @@ def _load_cached(self, desc:str, hash:str, sanitized_name:str) -> bool:
def _rank(self, image_features: torch.Tensor, text_embeds: torch.Tensor, top_count: int=1, reverse: bool=False) -> str:
top_count = min(top_count, len(text_embeds))
text_embeds = torch.stack([torch.from_numpy(t) for t in text_embeds]).to(self.device)
with torch.cuda.amp.autocast():
with torch.amp.autocast(device_type='cuda' if self.device == 'cuda' else 'cpu'):
similarity = image_features @ text_embeds.T
if reverse:
similarity = -similarity
Expand Down