Skip to content

[big tensor] fix median hang #74360

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

Merged
merged 2 commits into from
Aug 4, 2025
Merged

[big tensor] fix median hang #74360

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3e91d28
[big tensor] fix median hang
DongBaiYue Aug 1, 2025
92a97cc
fix coverage
DongBaiYue Aug 1, 2025
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
18 changes: 17 additions & 1 deletion python/paddle/tensor/stat.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -628,7 +628,23 @@ def median(
axis += dims
sz = x.shape[axis]
kth = sz >> 1
tensor_topk, idx = paddle.topk(x, kth + 1, axis=axis, largest=False)
# Use `sort` when:
# 1. The axis is not the last dimension (memory non-contiguous)
# 2. The axis size exceeds 10000 (heuristic threshold for performance crossover)
# Rationale:
# - `paddle.topk` in non-contiguous dimensions has O(N*k) complexity (k=n/2 for median → O(n²)). in paddle/phi/kernels/gpu/top_k_kernel.cu
# - `paddle.sort` has guaranteed O(n log n) complexity regardless of axis
use_sort = (axis != dims - 1) and (sz > 10000)
if use_sort:
sorted_x = paddle.sort(x, axis=axis, stable=True)
tensor_topk = paddle.slice(
sorted_x, axes=[axis], starts=[0], ends=[kth + 1]
)
if need_idx:
idx = paddle.argsort(x, axis=axis, stable=True)
idx = paddle.slice(idx, axes=[axis], starts=[0], ends=[kth + 1])
else:
tensor_topk, idx = paddle.topk(x, kth + 1, axis=axis, largest=False)
if mode == 'avg':
dtype = (
'float64'
Expand Down
Loading