Skip to content

guard torch.cuda.synchronize for CPU/MPS eval paths #271

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

guard torch.cuda.synchronize for CPU/MPS eval paths #271

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
e92d672
guard torch.cuda.synchronize for CPU/MPS eval paths
LiudengZhang May 15, 2026
a22a2bf
Use device_type variable for CUDA guard consistency
LiudengZhang May 15, 2026
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
5 changes: 3 additions & 2 deletions src/state/_cli/_emb/_eval.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -118,8 +118,9 @@ def load_config_override(config_path: str | None = None):
with torch.no_grad():
with torch.autocast(device_type=device_type, dtype=precision):
for batch in tqdm(dataloader, desc="Processing batches"):
torch.cuda.synchronize()
torch.cuda.empty_cache()
if device_type == "cuda":
torch.cuda.synchronize()
torch.cuda.empty_cache()

# Compute embeddings
_, _, _, emb, ds_emb = model._compute_embedding_for_batch(batch)
Expand Down
3 changes: 2 additions & 1 deletion src/state/emb/nn/eval_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -76,7 +76,8 @@ def evaluate_de(model, cfg, device=None, logger=print):
pred_exp = model._predict_exp_for_adata(
tmp_adata, cfg["validations"]["diff_exp"]["dataset_name"], cfg["validations"]["diff_exp"]["obs_pert_col"]
)
torch.cuda.synchronize()
if torch.cuda.is_available():
torch.cuda.synchronize()
Comment on lines +79 to +80
Copy link
Copy Markdown

@gemini-code-assist gemini-code-assist Bot May 15, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

This synchronization is likely redundant. Since _predict_exp_for_adata returns a result that is subsequently used as a DataFrame or numpy array (e.g., in compute_gene_overlap_cross_pert), the data transfer from the GPU to the host has already performed an implicit synchronization. You may consider removing this block entirely to simplify the code.

de_metrics = compute_gene_overlap_cross_pert(
pred_exp, true_top_genes, k=cfg["validations"]["diff_exp"]["top_k_rank"]
)
Expand Down